tecord.1132

	LIST OF TECO COMMANDS, TECO VERSION 1132
	Last updated 5 August 1981.  Z=230457

 (Note that a caret followed immediately by another character
 signifies a control character, except inside FS flag names, where
 that is not necessary because control characters are not allowed.)

^@:
<n>^@	for nonnegative <n>, is the same as ".,.+<n>".
	For negative <n>, is the same as ".+<n>,.".
	"10^@XA" puts the 10 characters after the pointer
	in a string in qreg A.
	<n>,0^@ returns -<n>;  this is often useful.
	<m>,<n>*0^@ returns -<m>;  it undoes the ",<n>".
	In other words, it extracts the first of two args.
	If you want <m> instead of -<m>, use another ,0^@.
<m>,<n>^@
	returns the value <n>-<m>.
<m>,<n>:^@
	returns <n>,<m>.

^A	inclusive-or (an arithmetic operator).
	^A may be used in a file specification as the first or
	the second filename to stand for the default first name.

^B	is a command for cleaning up after failing searches.
	If the last search was successful, ^B does nothing.
	If the last search failed, ^B moves to the end of the
	range searched if it was a forward search; to the beginning,
	if it was a backward search.  :^B does the same thing for
	failing searches but does FKC after successful ones.
	Note: ^B inside search string is a
	special char which is matched by any
	delimiter character.  The set of delimiter chars is
	specified by the contents of q-reg ..D;  initially, the
	delimiter characters are precisely the non-squoze
	characters (that is, all except letters, digits, ".", 
	"%" and "$").
	^B may be used in a file specification as the first
	or second file name to stand for the default second name.

^C	when typed in from console, terminates the
	command string, and starts execution.  If
	the command executes without error, TECO
	returns to its superior without flushing the type-in
	buffer.  When proceded, it will automatically
	redisplay the buffer on display consoles.
	When TECO returns, AC 2 will contain
	the address of the 7-word "buffer block" describing the current
	buffer - see the section "buffer block" at the end.
	To type in a ^C in a TECO command string, use ^]^Q^C,
	which is specially arranged to inhibit the normal
	action of ^C at command string read-in time.
	A ^C encountered as a command is an error.

^F	inserts its string argument, after deleting the last
	thing found with an S search or inserted with I or \
	(won't work if pointer has moved since the S, I or \
	was done).  Precisely, ^F is the same as FKDI.

^G	causes a "quit" by setting FS QUIT to nonzero.
	The consequences of that depend on the value 
	FS NOQUIT.  Normally, FS NOQUIT is 0;  ^G will then
	stop whatever TECO is doing and return to its top-level
	loop, or to the innermost ^R invocation if any, to
	read more commands (but first TECO will
	redisplay the buffer).  In particular, it will cancel a
	partially typed-in command string.
	All unread input except the ^G itself is flushed.
	On printing terminals, output is also discarded.
	(If FS NOQUIT is -2 or less, this is inhibited).
	If FS NOQUIT is positive,
	^G still sets FS QUIT but that has no effect.  Thus,
	a program can inhibit quitting temporarily, or quit
	in its own manner by testing FS QUIT itself.
	If FS NOQUIT is negative, setting FS QUIT nonzero
	causes an ordinary error (whose error code is "QIT"),
	which may be caught by an errset (:< - >).  If FS NOQUIT
	is less than -1, then in addition the input and output
	on the terminal are not	discarded.

^H	backspace;  it is illegal as a command.

^I	tab;  in bare TECO, it is a self-inserting character.
	In EMACS, or whenever FS ^I DISABLE is appropriately set,
	it is equivalent to a space.  In bare TECO, it inserts itself
	as well as the following string argument, so that Tab is
	equivalent to I Tab.  This was intended to be typed by the
	user himself;  using it in a macro is foolish.

	Tab characters in the buffer or in typeout are displayed
	using tab stops at a spacing controlled by FS Tab Width.

^J	line feed;  flushes current value.

^K<string>
	valrets a string argument to DDT with
	dollar signs replaced by altmodes.
	(To cause a dollar sign to be valretted, use
	"^]^Q$" (ctl-close ctl-Q dollar)).
	If the command string contains an P
	command, TECO command execution will
	continue with the character after the
	altmode ending the text string of the ^K.
	^K causes TECO to believe that the screen has
	been clobbered, so it will automatically clear the
	screen and redisplay everything at the next
	opportunity.  To avoid this, use "@^K" (if for
	example you know DDT will not type anything out,
	and will P the TECO).
	When TECO executes the .VALUE, AC 2 will contain
	the address of the 7-word "buffer block" describing the
	current buffer - see the section "buffer block" at the end.

^L	form feed;  clears screen on displays (when executed,
	not when typed).  See F+ for more details.

^M	carriage return;  flushes current value.
	If FS STEP MACRO is a string, it is executed (if there is a
	^J after the ^M, it is skipped over first).  If FS STEP MACRO
	is a nonzero number, ^M does these things designed to
	step through the program a line at a time:
	it displays the buffer unless there was typeout recently,
	then reads in a character and acts according to it.
	Most characters simply tell ^M to return so that more
	commands will be executed.  However, there are the
	following special characters:

	^F	quit.  Like ^G, but ignores the setting
		of FS NOQUIT and does a real quit.

	^P	end stepping.  Zeroes FS STEP MACRO and then
		proceeds without stepping.

	^R	enter ^R mode.  On return from ^R,
		another character will be read and decoded.

^N:
<n>^N	sets the FS LINES flag to <n>.  Like "<n>FS LINES".
	FS LINES controls the number of lines used for buffer
	display and, on display terminals, for all other output.
:^N	complements the FS TTMODE flag (initially 0).
	TECO normally displays the buffer on printing terminals
	only if this flag is set.  User buffer display macros
	should exhibit similar behavior.
:<n>^N	like "<n>^N :^N"
	Note: ^N in a search string is a special char which is
	matched by any char other than the char after
	the ^N in the search string.

^O<filename>
	bigprints <filename> on the device open for output.
	Note: ^O in a search string is a special character
	signifying "OR" i.e., it divides the search string into
	two strings either of which will satisfy the search.
	Thus, SFOO^OBAR will find either FOO or BAR, whichever is
	encountered first.

^P	alphabetic (ASCII) or numeric sort command.
	The entire buffer, or the part within the virtual boundaries,
	is sorted, after being divided into sort records
	(i.e., things to be sorted) on the basis of the arguments
	given to the command in the form of
	three TECO command strings following
	the ^P, separated by altmodes
	(Notes: (1) two successive null args
	will result in a premature end of
	command input, so use spaces where
	needed;  (2) a dollar sign in any
	arg will be replaced by an altmode;
	(3) the three args will be left in q-regs ..0, ..1, ..2).
	The three expressions are used to divide the buffer into sort
	records, each of which has a sort key.
	A sort key may be any substring of the record, or it may
	be a number.  The records and keys are found as follows:
	 1. The pointer is moved to the
	beginning of the buffer, which is the
	beginning of the first sort record.
	 2. The first command string is executed.
	If the key is to be a substring of the record,
	this command string should leave point at the
	beginning of the key.  If the key is to be a number,
	it does not matter what this command string does.
	 3. The second command string is executed.
	If this command string returns a value, that value is the key.
	Otherwise, this command string should leave point at the end
	of the key.
	 4. The last command string is executed.
	This should move the pointer to the end of the sort record,
	which is also the beginning of the next record.
	 5. If step 3 or 4 leaves the pointer
	at the end of the buffer, the creation of sort
	records is complete, and the sort
	takes place.  Otherwise, go back
	to step 2.
	Sort records and keys may be variable length.
	No character (i.e., a shorter key) sorts before ^@,
	and keys are considered left-justified for the comparison.
	Numeric keys are always "less" than any string keys.
	Failing search commands are not errors while scanning for sort
	records, unless the FS S ERROR flag is nonzero.  It is
	recommended that this flag be set, since code is easier to
	debug if all searches which are supposed to be allowed to fail
	have an explicit colon.

	There is nothing to prevent overlapping records
	from being specified;  the sort will copy each record
	so the overlap region will be duplicated.
	Insertion and deletion are allowed but know that
	TECO remembers the boundaries of records and keys as character
	numbers, so deleting chars from a record already delimited
	will shift chars from the next record into it, etc.
	The sort is stable.  :^P sorts in reverse order.
	If FS ^P CASE is nonzero, ^P ignores case;  that is,
	it sorts lowercase letters as if they were the
	corresponding uppercase letters.

^Q	^Q in a search string causes
	the next char to be quoted, i.e.,
	it is treated as an ordinary char
	even if it normally has a special
	meaning ("^Q^Q" is a normal ^Q;
	^Q works only at execution time, not at command string
	read-in time, so rubout cannot be ^Q'd).
	This also works inside file name specifications.

		^R MODE

^R	real time edit feature, intended mainly for display terminals.
	The position of the pointer is represented by the terminal's
	hardware cursor, rather than by any printed characters (^R
	ignores the contents of ..A, except on printing terminals).

	All non-control-non-rubout characters
	are normally self inserting;  the others are normally
	editing commands.  The user may redefine any character
	by means of the FS ^RCMAC flag.
	In ^R mode echoing is turned off, so typed-in characters
	manifest themselves only by their effect on the displayed
	buffer contents (but see FS ^R ECHO).

	Any command may be given a numeric argument, which most
	commands (including all characters that insert themselves)
	treat as a repetition count.  If no argument is specified,
	1 is the default, but commands can distinguish between
	an explicit 1 and a defaulted 1.  The argument is computed as
	follows:  <arg> = <basic arg> * (4 ** <exponent-of-4>)
	where <basic arg> is the explicit argument, if any, or
	1 otherwise.  An explicit argument is given with ^V or
	by control-digits.  <expt-of-4> is initially 0
	and incremented by ^U.  All commands except argument-setting
	commands discard their arguments even if they don't use
	the arguments.  Three flags contain the argument data:
	FS ^R ARG contains the explicit argument, if any, else 0;
	FS ^R EXPT contains the exponent of 4;
	FS ^R ARGP, if zero, indicates that no arg has been
	 specified (neither the explicit arg nor the exponent of 4);
	 if 1, indicates that only an exponent of 4 has been
	 specified, and the basic arg is still 1;
	 if 3, indicates that an explicit arg has been specified.
	 The 4 bit, if set, indicates that the argument should
	 be negated (Control-Minus sets this bit)
	All three are zeroed after any command that doesn't identify
	itself as an argument setting command by clearing FS ^R LAST.

	Any character may have a
	program associated with it, using the FS ^RCMACRO command.
	If that is done, when that character is typed, TECO
	will execute the program instead of inserting the char
	or using it as a built-in command.  The definition of a
	character may also be treated as a q-register in the
	"Q", "U", "X", "G", "[", "]", "M" and "FQ" commands;
	see "Q" for directions.  When the program is executed,
	q-reg ..0 will contain the character being handled.

	When errors take place inside ^R, or in macros called
	from ^R, after printing the error message TECO returns
	control to the innermost invocation of ^R (unless
	FS *RSET or ..P is nonzero).  The same thing happens
	for quits.  See ..P.

	One may wish to have
	a mode in which most editing commands are disabled, and
	most characters that are normally editing commands are
	self-inserting instead.  The FS ^RSUPPRESS flag, when
	nonzero, suppresses all built-in commands except rubout
	and all user defined commands whose definitions do not
	begin with "W" (since "W" at the beginning of a macro
	is a no-op, the only reason to have one there is to
	prevent suppression).  When a character is suppressed
	as a command, it becomes self-inserting.  An additional
	feature is the FS CTL MTA flag;  when it is negative,
	all control-meta-letters (and ctl-meta-[, ], \, ^ and _)
	have their definitions suppressed;  this mode is useful
	when editing TECO commands.

	In "replace mode", printing characters overlay a
	character instead of making the line longer.
	Replace mode is controlled by FS ^R REPLACE, which
	see for more details.  A more general facility for changing
	what ALL printing characters do is FS ^R NORMAL.  If this is
	nonzero, it counts as the definition of all "normal"
	characters.  That is, the definition which all printing
	characters initially have says to do whatever FS ^R NORMAL
	directs, and otherwise (if it is 0) insert or replace.

	You can redefine all "normal" characters whose Lisp syntax in
	..D is ")" by putting a function in FS ^R PAREN.  This
	function will be executed IN ADDITION to the actions for
	handling all "normal" characters.  This is good for something
	to point out the location of the matching open-parenthesis.

	Normal characters which you don't redefine in one of those
	ways will sometimes be subject to system echoing, when that is
	implemented.  This improves efficiency so it is done whenever
	it is safe.  Characters which have been redefined and thus
	made not normal cannot be echoed by the system.  If you
	redefine what "normal" means by making FS ^R NORMAL nonzero,
	system echoing will be turned off entirely.  Sometimes Space
	is redefined with a definition which never does anything but
	insert a space unless the horizontal position is beyond a
	certain point (at which point it will typically break the
	line).  Then it is desirable to enable spaces to be echoed
	when the cursor has not reached that column.  This can be done
	by means of the FS ^R EC SD flag.  Set it to the definition
	which Space has in that case, and set FS ADLINE to the column
	up to which it is safe to echo Space.  Then, when Space
	actually has that definition, the system will be told to echo
	spaces but only up to that column.


	By setting FS ^R INHIBIT nonzero, you inhibit display
	updating.  But ^R still remembers what is necessary,
	and as soon as FS ^R INHIBIT is zero again the
	updating will be done.

	Conversely, you may wish to inhibit the feature which causes
	redisplay to be suspended when input is available.  Setting FS
	D FORCE nonzero does this.

	The ^R-mode input dispatch table is actually indexed by
	9-bit TV character code.  Each 9-bit code can be redefined.
	The list of ^R-mode initial definitions that follows
	refers to the characters obtainable on non-TV's - in other
	words, the 9-bit characters which are the results of
	reading in the 14-bit codes 0000 through 0177, which are
	precisely the 9-bit characters which are equivalent to
	some 7-bit ASCII character.
	A subsystem which is not TV oriented need not worry about
	the 9-bit character set;  by using FI, and FS ^RCMACRO
	always without the atsign modifier, it can handle ASCII
	characters throughout.  TECO will automatically do the
	conversion to and from 9-bit characters on TV's.
	For those who wish to handle the 9-bit character set,
	the definitions of all 9-bit characters
	are listed in the section "TECO's character sets",
	along with the appropriate conversions between character
	sets.

	One may wish to have some operation (such as filing
	the buffer away) performed every so often while in ^R
	mode.  See "..F" for how to do this using the
	"secretary macro" feature.  FS ^R DISPLAY can be set
	to a macro which will be run every time ^R is about
	to do nontrivial redisplay.

	Although ^R mode is intended for display terminals,
	the creation of large macro-systems intended for use
	with ^R mode has made it necessary for ^R to work
	at least marginally on printing terminals.
	Since the physical cursor is not suitable, the ordinary
	TECO cursor is used (whatever is in ..A).  The buffer
	is displayed only when the screen is "cleared", such as by
	giving the built-in ^L command.
	Also, unless FS ^RECHO > 0, characters actually read
	by the ^R-mode command loop are typed out, although
	echoing is still turned off.
	This echoing can be made to happen even on displays
	by making FS ^RECHO negative (this is unwise to do
	if there is no echo area).

	Setting FS ^R SCAN to nonzero causes ^R commands to
	try to imitate printing terminal line editors by
	echoing the characters that they insert/delete/move over.
	In this case, FS ^R ECHO should be set to 1.
	FS ^R MAX is the maximum size deletion or insertion
	which ^R will try to print in the fancy way.
	If you define FS ^R TTM1, it gets to try to handle any
	redisplay ^R does not have a way of handling.  See its
	definition.
	Note that ^R recognizes insertion, deletion, and cursor
	motion from the values returned by the user's command
	definitions - see the following paragraph.

	Macros and ^R - reducing redisplay:

	Whenever control passes from normal TECO to ^R
	(that is, when a ^R is executed, when a @V is executed
	within a ^R-mode macro,	or when a ^R-mode macro returns),
	^R must be able to update the screen according
	to the changes that have been made in the buffer
	since the last time ^R mode lost control.  ^R can
	do that in a way that makes no assumptions, but
	that way is slow.  If information is still available on
	what areas of the buffer were changed, that info
	can be passed to ^R in the form of numeric args,
	and ^R will save time by assuming the info to be
	correct.  If the info is not correct, the screen
	will not be properly updated.  The options are:
	no args - the usual case - means assume nothing.
	One arg means that the buffer has not changed,
	although the pointer may have moved.  The actual
	value of the arg does not matter in this case.
	Two args should specify a range of the buffer
	outside of which nothing was changed.  ^R will
	limit redisplay to that range if possible.
	^R also knows what to do about macros that type text
	out;  if Q..H is nonzero when ^R is entered or
	returned to, ^R will not do any displaying until it
	has read one character (and executed it, unless it
	is a space).  If the selected buffer when you return to ^R is
	not the same one that was selected when ^R was previously in
	control, then it doesn't matter what args you give.

	If you like ^R mode, try:
	:I..G EL 90^S ^R :I..B Q..H"N 90^S ' ^R

	The commands are:

	 Control-Minus
		negates the argument of the next command.

	 Control-digits
		accumulate a numeric argument for the next
		command.  Thus, control-5 ^N will move down
		five lines.

	 ^A	go to beginning of current line (0L).
		With argument, <arg>-1 L.

	 ^B	go back over previous character (R)

	 ^C	complements the state of the comment mode switch.
		Types "C" for comment or "T" for text at the bottom
		of the screen, to say what mode you're in.
		When in comment mode, the ^N and ^P
		commands begin by going to the end of the line and
		if the last character is a semicolon,
		deleting it and any preceding tabs.
		Then, after moving to the next or previous line,
		if the line has a semicolon in it the pointer
		will be left after the semicolon;  otherwise
		the pointer will move to the end of line,
		and enough tabs will be inserted to move
		the pointer at least to the specified comment column,
		followed by a semicolon.
		Numeric argument is ignored.

	 ^D	deletes the next character after . (D)
		If FS RUBCRLF is nonzero, ^D before CRLF deletes
		both the CR and the LF.

	 ^E	moves to end of line (:L).  With argument, <arg>:L.

	 ^F	goes forward over the next character (C)

	 ^G	flushes any numeric argument or case-shift,
		unsets the mark if it had been set,
		and resets the case-lock.
		When ^R is actually in control (as opposed to a
		macro running inside ^R), ^G's quitting action
		is suppressed, and ^G acts as a command instead.
		Thus, it does not flush any type-in.

	 ^H	(backspace) inserts itself.

	 ^J	(linefeed) inserts itself.

	 ^K	kills to eol (K).  With arg, <arg>K.
		The text deleted is put in q-reg ..K.

	 ^L	redisplays the screen (used to recover from
		datapoint lossage).  Chooses a new window.
		A numeric argument specifies the number of lines
		of buffer to display - useful on printing terminals.
		On displays, if only a part of the screen is being
		used at the moment, only that part is cleared.

	 ^M	inserts a carrage return-line feed.

	 ^N	goes to next line (L).  With argument, <arg>L.

	 ^O	inserts a CRLF, then backs over it.
		"^Ofoo" is equivalent to "foo^M" but
		often requires less redisplay.
		With argument, inserts <arg> CRLFs
		and backs over the last.
		If you want to insert several lines in the
		middle of a page, try doing ^U^U^O before
		and ^U^U^K afterward.

	 ^P	goes to previous line (-L).  More generally, -<arg>L.

	 ^Q	inserts the folling character directly,
		regardless of its meaning as a command.
		If the char isn't already in the input buffer,
		^Q will prompt with a "Q" at the bottom of the screen.
		An argument to ^Q causes it to insert the same
		character <arg> times.  ^Q is not affected by
		replace mode;  the quoted character is always
		inserted.

	 ^R	causes the column the pointer is at to become
		the comment column.  Argument is ignored.

	 ^S	reads a character and searches for it.
		"^SA" in ^R mode is the like "SA" in TECO.

	 ^T	sets the ^R-mode mark at the current pointer
		position.  The mark is really the value of
		FS ^RMARK and is used by the ^X and ^W commands
		in ^R mode.  If FS ^R MARK holds -1 there is no
		mark;  that is the case initially and after any
		insertion, deletion or quit in ^R mode.
		Attempting to use the mark when there is none
		rings the bell.

	 ^U	increments the exponent-of-4 for the next command.
		This usually is the same as repeating it 4 times.
		Does not use any previous argument, but leaves
		it around for the next command.

	 ^V	sets the basic arg for the next command.
		The argument is composed of digits optionally
		preceded by a minus sign, echoed at the bottom
		of the screen and turned into a number in the
		current radix (FS IBASE).  The first non-digit
		terminates the arg and is treated as a command.
		^G will flush the argument.

	 ^W	kills everything between the current pointer
		position and the mark, putting the deleted text
		in q-reg ..K.  If there is no mark, nothing is
		deleted and the bell is rung.

	 ^X	sets the mark at the current pointer position,
		and moves the pointer to where the mark had
		been;  in other words, exchanges the mark and
		the pointer.  Does nothing if there is no mark.
		Do this several times to see both ends of the
		range that a ^W command would delete.

	 ^[	(altmode) terminates edit

	 ^]	reads a q-reg name and executes that q-reg
		as a macro.  The q-reg should contain ordinary
		TECO commands, not ^R mode commands.  The numeric
		arg to the ^] will be given to the macro which
		will see it as the value of ^Y (If no argument is
		specified, ^Y will be 1, but F^X will indicate that
		the macro had no argument).  The macro may
		return values to ^R telling it which areas of
		the buffer may need redisplay (see below).
		If the macro is to return values, it should end
		with a space - otherwise, the values might get
		lost within TECO.
		Example: " .,( G..K .) "
		gets q-reg ..K and returns 2 values limiting
		the range of the buffer in which changes took
		place.

	 ^?	(rubout) deletes bacwards (-D).  If FS RUBCRLF
		is nonzero, rubout when the pointer is after a
		CRLF deletes the whole CRLF.

	 ctl-rubout
		deletes backwards like rubout, except that tabs
		are converted to spaces and the spaces are deleted
		one at a time.

^S:
<n>^S	if <n> is positive, sleep for <n> 30ths of a
	second.  If <n> is negative, sleep until system
	run time (what FS UPTIME gets) = -<n>.

	Note: ^S in a search string is a special character
	which should be followed by a Lisp-syntax value
	and matches any character whose Lisp syntax in Q..D
	matches the specifed syntax value.
	For example, ^S( matches any character whose Lisp syntax
	is "(".

<n>:^S	sleep for at most <n> 30ths of a second, returning
	immediately if there is any input available.
	Returns the value of FS LISTEN (nonzero if
	input is available).

^U	display in the user-specified manner the directory of
	the current default device.  That is, invoke the user's
	buffer display macro if any;  otherwise on display consoles
	display in the standard manner, but do
	nothing on printing terminals.  These are the same actions
	TECO always takes at the end of any command string whose
	last command was an E-command.
	Note: if ^U is typed as the first character of a command
	string, it is executed immediately when read.

^V	pops the "ring buffer of the pointer".  ^V when the
	first character of a command string acts immediately,
	resetting the pointer to the value it had before the
	last time it was moved.  Successive ^V's will undo
	earlier changes of the pointer.  Up to 8 changes are
	remembered to be undone.  Motion caused by the use of
	^V in this manner does not get saved to be undone.
	^V not the first character typed is slightly
	different.  It pops the ring buffer into the pointer,
	and returns as its
	value the number that then remains on the top.  If that
	returned value is put in Q..I (which is what gets
	pushed on the ring buffer at the end of the command
	string) you can fool TECO's top level into thinking
	that the pointer was not moved by the command string
	that just finished, so nothing will get pushed back on
	the ring buffer (this is exactly what ^V as the first
	character typed does).  If TECO's top level is not in
	use, the program that is running must be hacked up to
	push explicitly on the ring buffer (using <n>^V)
	in order for anything to appear on it.
	If ^V attempts to jump out of the buffer, the pointer
	is not moved, but the ring buffer is popped.  A "NIB"
	error happens.
:^V	returns the value on the top of the ring buffer,
	without popping it or changing the pointer.
<n>^V	is equivalent to <n>FS PUSHPT.  It pushes <n> onto
	the ring buffer unless <n> equals whatever is at the
	top of the ring buffer.
<n>:^V	pushes <n> onto the ring buffer unconditionally.

^W	pops all the way to top level,
	exiting from any break-loops and not running the user
	defined error handler in ..P.

^X	only defined inside macro.  Its value is the
	first arg of the M command which called the macro.
	See the F^X command for a more sophisticated
	way for macros to examine their arguments.
	Note: ^X typed as the first character of a command
	tells TECO to type out the whole error message
	associated with the most recent error.  If the flag
	FS VERBOSE is zero (normally true on printing terminals)
	TECO normally types only the 3-letter code.  Use ^X
	to see the whole message if you don't recognize the code.
	Note: ^X in search string is a
	special char which is matched by any
	character.
	^X may be used in a file specification as the first or
	the second filename to stand for the default first name.

^Y	like ^X, only second or only arg of the M command.
	If ^Y is the first char	typed in in a command string,
	the most recently typed command string longer
	than 7 characters (not counting the 2 altmodes)
	is inserted in the buffer.  This is a loss
	recovery procedure.
	^Y may be used in a file specification as the first
	or second file name to stand for the default second name.

^Z	normally causes an interrupt to DDT when typed.
	However, one can be given to TECO by quoting it
	with ^_ , in which case it is a normal command:
	with no arg, its value is a pseudo-random number.
<n>^Z	inserts <n> random letters before the pointer.

Altmode	terminates following text argument to
	certain commands;  two successive altmodes terminate a command
	string in the bare TECO command loop, and begin execution of
	the commands in the command string.
	Execution of an altmode as a command depends on the
	setting of FS NOOP ALTMODES.  If the flag is >0
	(old-fashioned mode), altmode acts like the ^_ command.
	If the flag is negative (default mode), altmode is
	a no-op.  If the flag is zero (losing mode),
	altmode is an error as a command.

^\	exits from the innermost macro invocation, unwinding
	the q-register pdl to the level it had when the macro
	was entered, and popping all iterations that started
	inside that macro.  Note that if Q..N is popped this
	way, it's previous contents (before the pop) will be
	macroed (after the pop is done).  This enables macros
	to arrange arbitrary actions to be performed whenever
	the macro is exited, no matter for what reason.
	See also FS BACK RETURN.
:^\	exits from the innermost macro invocation, without
	unwinding the q-register pdl.  It does pop iterations.
		^] string substitution

^]	is not really a command.  It is a special character
	that makes it possible to substitute the contents of
	a q-reg into the TECO command stream at any point (such as,
	inside an I or S command).  ^] is processed when
	TECO reads a character from the command buffer
	(ie.  Before anything like insertion or execution
	is done	to the character.).  It gobbles the
	next character and decodes it as follows:

	 ^A	sets the one-character flag (see below)
		then reads another character and
		interprets it as if it had been typed
		after a ^].

	 ^Q	gobbles another character and returns
		it to TECO superquoted (i.e. It will
		not act as a text terminator, in a
		search string, it will have no special
		effect, etc.)

	 ^R	is the beginning of the name of a q-reg to
		be substituted.

	 ^S	cause the superquote flag to be turned on
		(see below) then read another character as in ^A

	 ^T	cause the delimiter flag to be turned off
		(see below) then read another character as in ^A

	 ^V	followed by a q-register name, causes the char
		whose ASCII value is in that q-register as a
		number to be substituted in.  That is, after
		^^AU0, ^]^V0 will substitute an "A".

	 ^X	reads a string argument to the M command that
		called the current macro, and substitutes it in.
		^]^X pushes the current command buffer onto a
		special pdl, then causes the normal macro pdl
		to be popped one level (the macro pdl is
		pushed onto each time an M command is executed.
		It is also pushed onto by ^]<q-reg name> (see below)).
		TECO will then proceed normally, reading from
		what is essentially a string argument to the
		current macro, until an altmode is encountered.
		This altmode will not be passed to TECO, but will
		instead cause the command buffer to be repushed
		on the macro pdl and the special pdl to be
		popped, thus restoring the state of the world.
		If a real altmode is desired in a string 
		argument, ^]$ (dollar sign) should be used.
		If TECO had been in any state other than reading
		commands (i.e. Reading a string to be inserted)
		then the characters read in the string argument
		will be protected from being taken as text delimiters.
		Thus I^]^X is guaranteed not to terminate somewhere
		in the macro argument.  If this is for some reason
		undesirable, a ^T (see above) should be used between
		the ^] and  the ^X (^]^T^X).  Characters are
		not normally protected from being interpreted
		specially in searches, etc.  If this is desired,
		use ^S  (eg. ER^]^S^X bar will cause the file 
		<macro argument> bar to be selected for read,
		even if the macro argument has spaces,
		semicolons, etc. in it.).
		If the one character flag had been on
		only one character will be read as an argument
		instead of an entire string.

	 ^Y	acts like ^]^X, but only one character is taken
		from the previous command level.  Has the same
		effect as ^]^A^X.  Additional ^] calls will be
		chained through, with the final character com-
		ming from the last command level not indirected.

	  (altmode) pass a superquoted altmode
		back to TECO (same as ^]^Q )

	 ^]	pass an actual ^] to TECO

	 "	pass a superquoted doublequote back to TECO.

	 $ (dollarsign) pass an ordinary
		altmode back to TECO (see ^X above)

	 .	is the beginning of a q-reg name.
		Multi-character q-regs such as Q..A can be substituted
		with ^] just like single-character q-regs.

	 0-9	the current command stream (executing macro) is saved
		and TECO begins reading command characters from the
		contents of the specifed q-register 0-9.  After all of
		the characters in that q-register, which must contain
		a string or buffer, have been read, reading of the
		macro which contained the ^] itself will resume.
		Delimiter protection and superquoting are handled as
		for ^]^X.  For example, I^]1 is equivalent to G1
		(Note:  G1 is optimized to be much faster than I^]1),
		because any altmodes in the string in q-register 1 will
		be delimiter protected so that they do not terminate
		the I command.  S^]1 will search for whatever string
		is in Q-register 1.  Altmodes in the string will not
		end it, but ^B, ^N, ^O, etc. will have their normal
		special functions.  S^]^S1 superquotes the string and
		makes even those characters behave like ordinary ones.
		The contents of the q-register are saved so that
		storing into the q-register while they are being read
		does no harm.

	 @	@ ("indirect") causes the characters substituted in
		by the ^] to be treated as if they in turn had a ^]
		in front of them.  Thus, after :IA.B, ^]@A will
		substitute q-reg .B.  After :IA.Bfoo, ^]@A will
		substitute the contents of .B, followed by "foo".
		I may change that if I can see an easy way.

	 A-Z	like 0-9 (insert q-reg)

^^<char>
	(ctl caret) has the value of the 7-bit ASCII
	code for <char>.

^_	(note that in order to type this character to a
	program, it must be typed twice, due to ITS hackery)
	ends execution of the command string "successfully";
	the TECO will log out if disowned, or return to its
	superior if a ^C ended the typed-in command string.
	Otherwise, or after TECO is P'd, TECO will reset all
	stacks (if FS *RSET is 0), then maybe display the
	buffer or dircetory (using the user's supplied macros
	in Q..B and Q..G if any), and go on to read another
	command string.
	It is not wise to use this as a nonlocal exit from
	a macro;  that is what F< is for.  The main use is
	to restart TECO's command reading loop at the current
	stack levels - useful when a user-defined error handler
	wants to transfer to a TECO break loop.  TECO's command
	loop puts a ^_ at the end of every command string to
	make sure that it gets control back when the command
	string terminates.  Otherwise, in a break loop, control
	would return right back to the suspended program.

  (Space)
	Spaces around arithmetic operators are ignored.
	Spaces between two values in lieu of an arithmetic
	operator count as a "+", except that spaces by themselves
	before a command that can use an argument do not count
	as an argument.

!<label>! defines <label> for use by
	O command (q.v.).
	This contruct is also the standard way of putting
	comments in TECO macros.  It is ignored in execution.
	However, quotes and angle brackets inside the label
	are counted when quotes are being balanced in a conditional or
	angle brackets in an iteration.  This is useful, because if
	the commands in a conditional need to have unmatched quotes
	you can put in matching quotes inside a label to make the
	whole thing balance.  The same is true for unmatched brackets
	in an iteration.  For example:
	  Q1"N  !"! FTYou're losing'  or  < Q1; EDFOO < !>! >

"	starts a conditional.  The character after the " gives
	the condition.  It is followed by conditionalized
	commands, up to a matching '.  If an else-clause is
	desired, the ' should be followed immediately by "#,
	with perhaps CRLFs, spaces, or comments (see "!") in between,
	followed by the contents of the else-clause, followed by
	another '.  A conditional may return a value.
	The conditionalized commands must be balanced as to quotes.
	If some commands contain unmatched quotes, place extra quotes
	inside a label (see "!") to balance them out.

	The argument to the conditional is normally gobbled up by
	the conditional, and the first conditionalized command
	receives no argument;  see F" for a variant conditional
	that passes the argument along instead.
	FS STEPMAC will be processed on entry to a successful
	conditional.

	The conditions that now exist are:

	Char:	Condition succeeds if numeric arg to " is
	 A	the ASCII code for an alphabetic character.
	 B	the ASCII code for a delimiter character
	 C	the ASCII code for a non-delimiter character.
	 D	the ASCII code for a digit.
	 E	zero.
	 G	positive.
	 L	negative.
	 N	nonzero.
	 U	the ASCII code for an "upper-case" character
		  (anything other than 140 through 176 octal).

	Any condition can be reversed with a ":";  0:"E fails.

	The delimiter characters are those characters which
	are specified as delimiters by the contents of q-reg
	..D.  Initially, q-reg ..D is set up to specify that all
	non-squoze characters are delimiters, but the user can
	change that by setting q-reg ..D.
	Squoze characters are letters, digits, ".", "%" and "$".

	Conditionals operate by skipping the text up to the
	matching ' if they fail, and doing nothing if they
	succeed.  If the ' terminating a failing conditional is
	followed by "#, they will be skipped as well.  If the
	conditional succeeded, they would be executed - and "#
	is really a conditional that always fails.

	A special kind of conditional that returns a success or
	failure code can be obtained by putting a ' after the ",
	as in "'E.  This kind of conditional does not skip any code.
	Instead, it just returns -1 for "success" and 0 for "failure".
	Such conditionals are useful because it is easy to and or
	or them together and then test the result.
	For example, an expression whose value is the signum of
	the number in q-reg 0 is:  Q0"'L-(Q0"'G)

#	exclusive or (an arithmetic operator).

$	(dollar sign) the old lower-case edit mode:
	"-1$" is the same as "-1F$/" (first dollar, then altmode)
	"0$" is the same as "0F$" (first dollar, then altmode).
	"1$" is the same as "1F$".  For more info, see the
	"F$" command (that's dollarsign, not altmode).

%<q>	increments the number in q-reg <q> by 1,
	and returns the result as a numeric value.
	Meaningless if the q-reg contains text.

&	logical and (an arithmetic operator)

'	terminates a conditional (see ").
	This character is actually a no-op when executed.
	It is for the " to search for if the condition fails.

(,)	fill usual role of parentheses in arithmetic calculations.
	However, they are more general in that not merely arithmetic
	operations but arbitrary commands can be enclosed within them.
	The value within the parentheses is simply the value returned
	by the last command before the close.  If the open was
	preceded by an arithmetic operator, arithmetic is immediately
	done on that value.  Otherwise, the close merges the values
	from inside the parentheses with the values saved by the open.
	Thus, 1,(2) and 2(1,) are both equivalent to 1,2.
	The colon and atsign modifier flags are also merged in that
	the open saves them and clears them, and the close restores
	and merges them:  they will be on after the close if either
	they were on before the close or they were on after the open.

	Parentheses can also be used where a q-register name is called
	for; the value returned by the last command before the close is
	used as the "contents" of the q-register.  This construct is
	allowed only with commands that only examine the q-register,
	not with commands that store into it (since there is no place
	to store into).  See the Q command for this.
	See also F( and F) for variants of these commands.

*	multiplication (an arithmetic operator).
	Note that in TECO there is no operator precedence.
	Evaluation of arithmetic operators is left-to-right.
+	addition (an arithmetic operator).

,	separates arguments for commands taking two numeric arguments.
	Doesn't affect the colon and atsign flags.

-	subtraction (an arithmetic operator).

.	equals the number of chars to left of the pointer.
		..n Q-registers

..0,..1,..2
	"^P" sort puts its 3 arguments into these q-regs.
	These q-regs are also used by "F^A".

..A	holds the string to be used to represent the cursor
	in standard buffer display.  Initially "/\" on displays,
	"^A^B" on Imlacs (looks like an I-beam), and
	"-!-" on printing terminals (of course, TECO's default is
	not to display the buffer on printing terminals unless
	FS TTMODE is set).
	In the cursor, backspaces always really backspace
	and all other control characters are treated as non-spacing
	characters.

..B	holds the user buffer display macro.
	After each command string whose last command was not
	an E-command, TECO does "normal buffer display", as follows:
	if ..B is 0, as it initially is, the default is:
	on graphics devices, do "standard buffer display";
	on printing terminals, do so only if FS TTMODE is set;
	otherwise do nothing.  For details of standard buffer
	display, see "@V".
	If q-reg ..B is nonzero, TECO simply macroes it.  Normal
	buffer display in this case consists of whatever that
	macro happens to do.
	Q-reg ..H and flags FS ERRFLG and FS ERROR will contain
	information about the command string that just ended.
	If either Q..H or FS ERRFLG is nonzero, there is text
	on the screen that should not be immediately covered over.
	The buffer display macro should check ..H and not display
	if it is nonzero.  FS ERRFLG need not be checked, since
	if -1, it will automatically cause all
	typeout on the first line of the screen to be ignored
	on displays.  This is the right thing if the buffer display
	macro doesn't wish to worry about errors.  If it is
	desirable to write on the first line and overwrite the
	error message, just zero FS ERRFLG.

..D	holds the delimiter dispatch table, which tells several
	commands how to treat each of the 128 ASCII characters.
	These commands are FW, FL, "B, "C and the special search
	character ^B.  The treatment of the character with ASCII
	code <n> is determined by the values of the characters
	in positions 5*<n>+1 and 5*<n>+2 in the delimiter
	dispatch table.
	The first of the dispatch characters says whether the
	character <n> is a delimiter.  The dispatch character
	should be " " for a delimiter and "A" otherwise.
	This dispatch character is used by FW, "B, "C and ^B.
	The second dispatch character describes the character's
	syntax in LISP.  The possibilities are "(", ")", "/", "'"
	"|", " ", "+" and "A".  Each says that the character <n>
	should be treated by FL and @FW as if it were an open, a
	close, a slash, etc.  "+" is treated the same as "A" by the FL
	command.  The distinction between "A" and "+" is there to be
	used by certain TECO programs.
	Initially, the first dispatch character is "A" for squoze
	characters (letters, digits, "$", "%" and "."), and
	" " for all others.  The second dispatch character is set
	up to reflect the default LISP syntax definitions as closely
	as possible.
	The delimiter dispatch must be at
	least 640 characters long so that every character has
	a dispatch entry.  ..D should always contain a buffer
	or a string;  if it holds a number an error will result.

..E	holds the output radix for = and \.  Initially decimal.
	Negative radices work - somewhat.  If the radix is 0 or 1,
	the next attempt to use it will change it to decimal
	and also cause an error "..E".

..F	holds the ^R secretary macro.  If nonzero,
	it will be macroed every (FS ^RMDLY) characters
	while ^R mode is in use.  More precisely, the counter
	FS ^RMCNT is decremented each time through ^R's main
	loop, and if it becomes 0, it is reset from FS ^RMDLY
	and ..F is macroed.  ..F is also macroed whenever the
	outermost level of ^R mode is exited (but not when
	inner recursive invocations of ^R are exited).
	When ..F is macroed because ^R is being exited, the
	FS ^RMODE flag will be 0;  otherwise it will be nonzero.
	If you are using ..F to save the buffer in a file,
	then when you temporarily bind the buffer or put strange
	crud in it, you should also bind ..F to 0 to prevent the
	buffer from being saved while it is meaningless.  Do not
	bind FS ^R MDLY, though;  then, if the time to run ..F
	comes up while in this state, TECO will make sure to run it
	as soon as it becomes nonzero again.

..G	holds the user-specified directory-display macro.
	Whenever TECO wants to display the directory in the
	usual manner (that is, when ^U or E^U is executed or
	at the end of a command string whose last command was
	an E-command), if this q-reg isn't zero TECO will simply
	macro it (otherwise, TECO has defaults - see "^U").
	When that is done, q-reg ..H will contain useful info.

..H	is the "suppress display" flag.  It is set to zero at the
	start of each command string, whenever the screen is
	cleared.  It is set nonzero when any typeout or display
	takes place, except for error message typeout.
	TECO's default is not to display the buffer if this q-reg
	is nonzero at the end of the command string.  User buffer
	and dir display macros should also look at this flag.
	If ..H is nonzero on entering or returning to ^R, ^R will
	wait until a character is typed in (and executed, unless
	it is a space) before allowing any redisplay.

..I	at the start of each command string,
	.'s value is saved in this q-reg.
	At the end of each command string, Q..IFS PUSHPT is done.
	Those actions are what enable the ^V command to work.

..J	initially 0, if this q-reg contains a string that string
	will appear on the screen just above the echo area, on
	the same line that --MORE-- sometines appears on.  The
	--MORE-- will still appear, following the ..J string,
	if it is appropriate.  The displayed string is not
	updated immediately when ..J is changed, but rather at
	the next opportunity for redisplay of the buffer or
	the next time typeout reaches the bottom of the screen.
	It is possible to put a buffer in ..J but that has the
	problem that TECO will not always be able to detect it
	when the buffer's contents change, and thus will not be
	able to update the screen when it should.
	If the string in ..J is too long, it is truncated to make
	sure that the --nn%-- or --MORE-- can fit.
	If recomputing ..J when you want it to change is too
	inconvenient, you can arrange for ..J to be recomputed when it
	is about to be displayed.  See FS MODE CHANGE.

..K	each ^K or ^W command in ^R mode puts the deleted text
	in this q-reg so it can be reinserted if desired.

..L	whenever TECO is started, this q-reg is executed.
	Also, after an EJ, the macro loaded into ..L is run.
	When a TECO dump file is made with @EJ, ..L should
	contain a macro to do whatever must be done when the
	file is loaded back in.  However, since that macro would
	be re-executed if TECO were G'd afterward, it should
	replace itself with something innocuous that just
	resets the terminal-related flags.
	If a ..L macro can exit, it should do so with a ^_ to
	start up the TECO command reading loop.  A CNM error
	will occur if the ..L macro tries to return any other way.
	..L should no longer be used for resetting
	terminal-type-dependent options.  Use FS TTY MACRO for that.

..N	this q-reg is special in that whenever it is popped by
	automatic unwinding of the q-register pdl, the previous
	contents are macroed after the pop.
	Thus, it is possible for a macro to set up an
	action that will be performed when the macro is exited,
	no matter what causes it to be exited, by pushing Q..N
	and putting the commands for that action in Q..N.  For
	example,  [0 .U0 [..N :I..N Q0J   saves . in such a
	way that it will always be restored.  That string,
	unfortunately, has a timing error in that a ^G-quit
	after the [..N will find an inconsistent state.  The
	remedy is to use the FN command which is the same as
	"[..N:I..N":  [0 .U0 FN Q0J
	Within a macro that has already set Q..N up in this
	way, the easiest way to add another action to be
	performed is to append to ..N using
	:I..N^]^S..N<new-stuff>.
	Note that popping Q..N explicitly with ]..N does not
	macro it.
	If you wish explicitly to pop ..N and macro the old
	value, the way to do it is "-FS QPUN".  "M..N]..N"
	has the disadvantage that when ..N is executed it is
	still on the q-reg pdl;  that may make it execute
	improperly and also is a timing error.

..O	this q-reg is defined to hold the current buffer.  That is,
	all the commands that use "the buffer" use whatever
	buffer happens to be in Q..O at the time.  An attempt
	to put a string or number in Q..O causes an error.

..P	holds the user-defined error-handler macro, if any.
	Whenever an error occurs that is not caught by an errset,
	this macro will be invoked.  If it is 0, TECO will
	instead print out the error message and set up for "?"
	in the normal manner.
	The executing command string will have been pushed on the
	macro pdl, so FS BACKTRACE can be used to examine it.
	Also, the arguments will have been saved with "(" so that
	they can be examined with ")F(=".
	FS ERROR will contain the error code for use with
	FE or FG in obtaining the error message.
	Note that the error handler is invoked for quits and when
	TECO is restarted, as well as after errors;  at those
	other times FS ERROR will be zero.
	If the error handler prints an error message in the main
	program area of the screen, it may wish to allow buffer
	display to occur as usual but prevent the error message
	from being overwritten by it.  Setting Q..H to zero
	permits buffer display, and setting FS ERRFLG to minus
	the number of lines of error message preserves them.
	The FG command takes care of this automatically.

	The error handler can return to the erring program with
	")^\" or "F)^\" (return whatever you like, but make sure
	to close the parentheses somehow).  However, do not expect
	the command that signalled the error to be retried.

	To abort the entire computation which got the error, the
	correct thing to do is FS ERR THROW.  This will return to the
	innermost error catch (:@< ...>) or ^R level.  If there is
	none of those, it will return to the top level of TECO.
	You may also use F; to return to a catch that was
	made at a higher level, or use ^W to pop out to TECO's
	top level loop.

	To make a break loop, just do a ^_ which will transfer
	to TECO's command string reader.

	If an error happens while the error handler is being entered,
	in order to prevent an infinite error loop TECO does an
	automatic ^W command to pop all stacks and prints "error
	entering error handler".  Stack overflow for any reason is
	likely to cause this since the error handler requires stack.
	Any error in the error handler is likely to cause a series of
	recursive errors, fill the stack, and cause this response.

..Q	holds a q-vector which serves as the symbol table for TECO
	variables, such as QFoo accesses.  The symbol table is
	in the format that FO likes.  Initially, the q-vector in
	..Q has only one element, which contains 2, the default number
	of words per symbol table entry.  See FO and Q for more
	details.

..Z	initially holds the same thing as ..O (the initial buffer),
	on the assumption that your main editing will be done in it,
	so that if you accidentally leave something else in ..O
	you can do Q..ZU..O to recover and not lose all your work.

/	division (an arithmetic operator).

0-9	a string of digits is a command whose value is a number.
	If it is not followed by a ".", the normal input radix
	(the value of FS IBASE, initially 8+2) is used.
	If the number ends with ".", the radix used is
	the value of FS I.BASE, initially 8.
	An attempt to type in a number too large for a 36-bit
	word to hold will cause a "#OV" error, unless the radix
	is a power of 2.

:	used before certain commands,
	modifies function of that command
	in a way described separately for each such command.
	Arithmetic operators and comma do not affect the
	colon flag.  Parens save it just like arguments, and
	don't deliver it to the commands inside the parens.
	Commands that don't return values always turn it off;
	commands that do, either ignore it or use it and turn it off.

;	does nothing if arg<0.  Otherwise
	sends command execution to char
	after next > (see < description).
	If no arg, uses value returned by
	last search (see S) as arg.
:;	like ;, but with the opposite condition:
	end iteration if arg is <0, or last search succeeded.
@;	like ;, but exit iteration if arg is EQUAL to zero.
:@;	like ;, but exit iteration if arg is NOT EQUAL to zero.

<	begin iteration.  Commands from here to the matching >
	are executed over and over, until a ; command decides to exit
	the iteration.  It does so by scanning forward, counting <'s
	and >'s seen, until an unmatched (since the ;) > is passed.
	Because of the way ; works, commands inside iterations
	following a ; must be balanced as to angle brackets.  If a
	command contains an unmatched bracket, you can use an opposite
	bracket inside a label to balance it out.
	Within the iteration, the sequence !<!> can be used to mean
	"jump to the beginning of the iteration."  The !<! is needed
	to avoid confusing any ; commands earlier in the iteration.
	It is an error if the iteration remains unterminated
	at the end of the macro level it began on.
	Within iterations, failing searches do not cause
	errors, unless FS S ERROR is nonzero (as it is in EMACS)
	to disable this "feature".
<n><	iterates <n> times (or until a ; command exits).
	The !<!> construct must not be used in such an iteration
	since, if the count has run out, it will not do its job.
	Use !<!@> if you want to jump to the beginning without
	decrementing the count.
<m>,<n><
	ignores <m>.

:<	begin errset.  This is like < except that errors
	occurring inside it are caught and will return after the >.
	TECO will not print any error message nor run the error
	handler in ..P.  Instead, the error message string will be
	returned as the value of the ">" which ends the errset.
	If the errset returns normally (no error) the value is 0.
	The canonical application for errset is 1:< EDFOO BAR > which
	deletes FOO BAR if it exists but avoids any problem if the
	file does not exist.
	TECO's built-in error messages all start
	with a 3-letter code that says what kind of error it is;
	use a F~ to compare the error message string against a
	specific prefix.
	Note also that :< iterates like <.  Perhaps you want 1:< ?
	Errsets do not prevent failing searches from erroring
	(luckily), and prevent any iterations outside them from
	doing so.  An iteration inside an errset will still prevent
	searches from erroring inside it despite the errset.
	Of course, to really win, you should set FS S ERROR.
	An errset has no power through a call to ^R or through an
	error catch (:@<).  The innermost errset, ^R, or error catch
	is what determines how an error will be processed.

:@<	begin error catch.  An error catch specifies a place for
	control to return to after an error has been processed (error
	message printed or error handler in ..P run).  It differs from
	an errset (:< with no @) in that the errset grabs control
	without letting the error message be printed.  The canonical
	application for an error catch is in a command loop which
	allows the user to give many different commands.  An error in
	one of the commands should be reported as usual, but the error
	should not cause the command loop to be exited.  Every call to
	^R implicitly contains an error catch.  Note that :@< iterates
	just like <.  If you want it to execute only once, do 1:@<.
	The command FS ERR THROW is a way to return control to the
	innermost error catch (where ^R levels count as error catches).
	The value returned by the error catch will be 0 if it was
	exited normally, by falling through the end.  If FS ERR THROW
	was used to exit, then the value will be the argument given to
	FS ERR THROW.

=	is for printing numbers:
<n>=	types out <n> in the current output radix, and a CRLF.
	The output radix is kept in q-reg ..E .  It is initially 8+2.
<m>,<n>=  types both <m> and <n>, with a comma between.
:=	is like = but omits the CRLF.
@=	is like = but types in the echo area.  @:= also works.

>	end of iteration, errset or catch (see "<", ":<", "F<").
@>	end of iteration, etc., except that the iteration count
	is not decremented.  !<!@> is a good way to jump back to
	the beginning of the iteration without decrementing the count.

?	if this is the first char input after
	typeout of an error message from TECO
	several command chars before the one
	causing the error will be typed.
	Otherwise, enter trace mode, or, if in trace mode
	already, leave trace mode.
	When in trace mode all command
	chars are typed out as they are executed.
	Trace typeout never uses the first line so that
	error messages won't wipe it out.
	The flag FS TRACE is nonzero when in trace mode.
:?	leaves trace mode whether in it or not.

@	is a modifier command which alters the action of
	the following command (except for a few commands which
	ignore it and pass it on to the next command).
	Many commands which take string arguments interpret
	the @ modifier to mean that the first character after
	the command should be used as the string delimiter,
	instead of Altmode.  But not all string argument commands
	do this, so you must check the individual command.
	The older form of the @ command is the ^ command,
	so most old TECO programs will contain ^ instead of @.
	^ was changed to @ because it could be confused with control
	characters.

A	if no arg, append next page of
	input file to current contents of
	buffer, i.e., like "Y" only don't empty buffer first.
	If virtual buffer boundaries are in use, the appended
	text goes just below the upper virtual boundary.
	Does not close the input file.
<n>:A	appends <n> lines of the file (but won't append beyond
	a page boundary).  Uses the same conventions for throwing
	away padding as "Y" does.  Does not close the input file.
@A	appends all the rest of the file.  A cross between
	"A" and "@Y".  Closes the input file.

<n>A	value is the 7-bit ASCII value of char arg chars
	to the right of the pointer.  Note that
	"0A" is the character immediately to the left of
	the pointer and "-<n>A" is the character <n>+1
	characters left of the pointer.  If .+<n>-1 is not
	within the bounds (real or virtual) of the buffer, a
	"NIB" error occurs.
<m>,<n>A
	is like <n>A except that when <n>A would cause a
	"NIB" error, <m>,<n>A will return <m>.  Thus,
	13,1A will return 13 iff the pointer is either at the
	end of the buffer or before a carriage return.

B	normally 0.  Actually, the number of the first character
	within the virtual buffer boundaries - but that will be
	the first char in the buffer (char number 0) unless you
	have used FS BOUND or FS V B to change that.

C	moves in the buffer relative to pointer:
<n>C	move pointer <n> chars to the right (1 char, if no arg).
	If that's out of the buffer, a "NIB" error results.

:C	like C, but returns -1 ordinarily,
	or 0 if C without colon would cause an error.
	:C is to C as :S is to S.

D	delete arg chars to right of pointer.
	If arg<0, delete to left of pointer.
		"E" commands

E	is the prefix for most operations on files.

E^U<dir>
	displays in the usual manner the directory of the
	device specified in the string argument, or the default
	device.  More precisely, reads the string arg and sets
	defaults, then does "^U".

E?<file>
	tries to open <file>, and returns 0 if successful.
	Otherwise, the value is the system error code which
	describes why the file cannot be opened.
	In any case, the file does not stay open, and the open
	input file if any is not interfered with.
	The file's reference date is not set.
	If the device is USR:, the job will not be created if it does
	not exist.

E_<old><new>
	makes a copy of the file <old> and names it <new>.
	I/O is done in ASCII block mode.  The currently open
	input and output files are not affected.  :E_ is similar but
	transfers the date of the input file to the output file.
	@E_ puts the real filenames of the input file into the
	defaults before reading the output file name, making it easy
	to preserve the version number.

EC	close the input file, if any.  This should always be done
	whenever an input file is no longer needed;  otherwise, one
	of the system's disk channels will be tied up.
	@Y, EE and EX automatically do an EC.
	All other input operations always leave the input file open.

ED<file>
	deletes <file>.

:ED	deletes the currently open input file (ITS only).

@ED<file>
	on Twenex, deletes and expunges <file>.
	On ITS, same as plain ED.

EE<file>
	like infinity P commands then EF<file> and EC.

EF<file>
	files output accumulated by PW and
	P commands with the name <file>.  <file> may not contain
	a device or SNAME;  they must have been specified when
	the file was opened (with EI or EW) and may not be changed.

:EF	files the output without renaming the output file.

EG	is a semi-obsolete command for getting random info.
	It inserts in buffer on successive lines
	the current date (as YYMMDD),
	the current time (as HHMMSS),
	TECO's current sname,
	TECO's default filenames for E-commands,
	the real names of the file open for input
	the date in text form,
	a 3-digit value as follows:
		1st digit = day of week today (0 = Sunday)
		2nd digit = day of week of 1st day of year
		3rd digit should be understood as binary:
			4-bit = normal year, and after 2/28
			2-bit = leap year
			1-bit = daylight savings time in effect.
	    (this line is blank on Twenex)
	and the phase of the moon.
	There are better ways to get most of these things:
	FS FD CONV, FS D FILE, FS D SNAME, FS I FILE.

EI	opens a file for writing on the default device.
	The filenames used will be "_TECO_ OUTPUT".
	When the output file is closed, it will normally be
	renamed to whatever names are specified.  However, if
	the TECO is killed, or another output file is opened,
	anything written will be on disk
	under the name "_TECO_ OUTPUT_"
:EI	like EI, but uses the current filename defaults
	instead of "_TECO_ OUTPUT".  This is useful for opening
	on devics which do not support rename-while-open
	fully, such as the core link.  Use :EF to close the file
	without renaming it.
@EI	like EI, but opens an old file in rewrite mode
	if there is one, rather than creating a new file
	in all cases.  Together with FS OFACCESS and
	FS OFLENGTH this can be
	used to update an existing file in arbitrary ways.
	However, what you really want to use is:
@:EI	like @EI but uses the default filenames
	rather than "_TECO_ OUTPUT".

EJ<file>
	restores the complete environment (q-reg values,
	buffer contents, flag settings, etc) from the
	specified file, which should be in the format
	produced by @EJ.  This restores all q-regs, buffers,
	and flags to what they were when the file was
	dumped.  Exception:  pure (:EJ) space is not changed,
	nor is FS :EJPAGE.  After loading, TECO restarts itself,
	which implies that if a nonzero value was
	loaded into Q..L, it will be macroed.
	This is intended to be used in init files, for
	loading up complicated macro packages which would
	take a long time to load from source files.
	If the file isn't a dump file, or was dumped
	in a different TECO version, an "AOR" error occurs.

@EJ<file>
	dump all variable areas of TECO on the file open
	for writing (it must already be open), and file it
	under the specified filenames.  One should not
	write anything in the file before doing "@EJ".
	Files written with @EJ can be loaded into a TECO with
	the EJ command, or they can be run as programs directly,
	in which case they will bootstrap in all the constant
	parts of TECO from the canonical place:
	.TECO.;TECPUR <TECO version>.  They also contain indirect
	symbol table pointers to that file, so that attempting to load
	the symbols of an @EJ'd file will load TECO's symbols.
	If you @EJ a file TS FOO on your home directory, then FOO^K
	will always get you that environment.

:EJ<file>
	inserts the specified file into core, shareable and
	read-only, and returns a string pointer to the beginning
	of it.  :EJ assumes that FS :EJPAGE points to the lowest
	page used by :EJ's, and inserts the file below that
	page (updating the flag appropriately).  Memory is
	used starting from the top of core and
	working down to page 340.
	See the sections "buffers - internal format"
	and "buffer and string pointers - internal format"
	for information on what can go in the file.
	An ordinary ASCII text file is not suitable for :EJ'ing.
	A file to be :EJ'ed must, first of all, be a single
	string whose length (including its header) must be a
	multiple of 5120 (1K words of characters).  Within that
	string lives the other strings or whatever that are
	the data in the file.  Their format is unrestricted
	except that the first thing in the file (starting after
	the header for the file as a whole) should be a string
	which is the file's "loader macro" which must know how
	to return the data in the file when asked for it.
	The loader macro should expect to be called with the
	name of the desired data (as a string) as the first
	argument (^X), and a pointer to the whole file (as a string)
	as the second argument (^Y).  The reason for passing it
	the pointer to the file is so that the loader itself can
	be pure (independent of the particular file containing it).
	The pointer to the file, plus 4, gives a pointer to the
	loader itself, if the loader wishes to examine its body.
	The loader macro should return as its value the string
	which is the value associated with the specified name.
	If the name is undefined in the current file, the loader
	should pass the request on to the loader in the next file.
	The next file can be assumed to start right after the
	end of the current one, so that ^Y+FQ(^Y)+4 is a pointer
	to it.  If there is another file, FQ of that will be
	positive;  otherwise (this is the last file in memory)
	FQ of that will be -1.
	If there are no more files, the loader should return 0.
	The goal is that several files with different loader macros
	should be :EJ'able in any order, and yet allow things to be
	loaded out of any of them at any time.

EL	display in the standard manner the directory of the
	default device.  This command
	does not use the user's buffer display macro;  in fact,
	the buffer display macro might well use this command.

EM	insert in buffer file directory of the default device.
	On Twenex, any pre-comma arg specifies don't include author name,
	and a post-comma arg is passed to the JFNS call.

EN<old><new>
	renames the file <old> to have the name <new>.
	The device and SNAME may not be changed;  they should not
	be specified in <new>.

EP<file>
	does ER<file>, then bigprints file name
	twice on device(s) open for writing.

EQ<from><to>
	creates a link named <from> pointing to the file <to>.
	devices COM:, TPL: and SYS: are understood.  (ITS only)
	An attempt to link to a non-disk device is an error.

ER<file>
	opens <file> for input.  The "Y", "A" and "FY" commands
	in various forms may be used to read from the file.
	As soon as the file is no longer needed (eg, if all
	of it has been read), an "EC" should be done to close
	the input channel.  "@Y" and "EE" do an automatic "EC".
	FS IF ACCESS, FS IF LENGTH, and FS IF CDATE make it
	possible to get or set various parameters of the file.
<n>ER	is similar but specifies open-mode <n> (actually,
	<n> IOR 2).  <n> should be even!!  <n>=4 specifies
	image mode, which makes a difference for file directories.
	This is only meaningful on ITS.
<n>,ER<file>
	bits in <n> are interpreted as follows:
	1: opens an input file and does not change its reference date.
	   This is defined to work on all operating systems.
	2: opens an input file in "thawed" mode.  This is probably
	   only useful on Tenex, where it allows multiple readers of
	   the file.
	4: (Twenex only) does not open the file at all, just does a GTJFN.
	   FS IF CDATE, FS IF NAME, and FS IF FDB can still be used, though.

ET<file>
	sets the default filenames to <file>.

:ET	(Twenex only) does a GTJFN reading from the terminal, to
	provide filename completion, etc.  The numeric argument is
	the left-hand flags for the GTJFN.  The flag FS :ET MODE
	controls which filenames' default values are provided for the
	GTJFN.  The pre-comma argument is the prompt string; on Tops-20
	this is used for setting up a call to the COMND JSYS.  On
	Tenex, it is simply typed out in the echo area; so this is
	the correct way to prompt when reading a filename.
EW<dir>
	like EI but device specified by
	following text string rather than by a numeric arg.
:EW<file>
	like EW, but uses the specified filenames
	instead of "_TECO_ OUTPUT".  This is useful for opening
	on devics which do not support rename-while-open
	fully, such as the core link.
@EW<dir>
	like EW, but opens an old file in rewrite mode
	if the file exists, rather than creating a new file
	in all cases.  Together with FS OFACCESS and FS OF LENGTH,
	this can be used to update an existing file in arbitrary ways.
	However, what you really want to use is:
@:EW<file>
	like @EW but allows filenames to be specified
	rather than using "_TECO_ OUTPUT".

EX<file>
	(Twenex only) if a file is open for output, does EE<file>,
	then instructs the EXEC to repeat the last CCL type command
	(load, execute, compile, debug).

EY<dir>
	like EL but specified device and SNAME.  On Twenex, the entire
	string, if provided, is used, including wildcards, for the GTJFN.
	Also, on Twenex, a numeric argument specifies the JFNS format.

EZ<dir>
	like EM but etc.

E[	push the input channel, if any.
	Saves the current input file and position in it, or saves
	the fact that no file is open.
	Useful for reading in a file without clobbering any
	partially read input file.
	Note:  for this and the next
	three commands ("E]", "E\", "E^"),
	the file open for input must be
	randomly accessible (=DSK).  The one
	open for output need not be.
	FS PAGENUM and FS LASTPAGE are saved by E[
	and restored by E].

E]	pop the input channel.
	If any input file was open, the rest of it is flushed.
	Further input will come from the file that was popped.
	(see "E[".)

E\	push output channel.
	(see "E[".)

E^	pop output channel.
	If an output file is open, it is closed without being
	renamed, so it is probably filed as "_TECO_ OUTPUT".
	(see "E[".)

		"F" commands

F	further decoded by the next character as follows:

F^@:
<m>,<n>F^@	
	returns 2 values, which are <m> and <n> in numerical
	order.  Thus, "1,2F^@" ans "2,1F^@" both return 1,2.
	"<m>,<n>F^@T" is the same as "<m>,<n>T" except that
	the former will never cause a "2<1" error.
<n>F^@	returns, in numerical order, 2 args that delimit a range
	of the buffer extending <n> lines from the pointer.
	Thus, "<n>F^@T" is the same as "<n>T".

<n>@F^@
	is like <n>F^@ but treats in CRLFs, not just CRs.

F^A:
<m>,<n>F^A<q>
	this command scans the range of the buffer from <m> to <n>
	using the dispatch table in q-register <q>.  That is, each
	character found in the buffer during the scan will be
	looked up in the dispatch table and the specified actions
	will be performed.  The dispatch table should be a string
	or buffer with at least 128*5 characters in it -
	5 for each ASCII character.  Each
	character seen has its ASCII code multiplied by 5 to
	index into the table, and the 5 chars found there are
	executed as TECO commands.  When that is done, the char
	that was found in the buffer is in Q..0 as a number,
	point is after the character if forward, before if backward,
	Q..3 holds the dispatch table that was in use (so that
	the dispatch commands can change it if they wish) and
	Q..2 holds the end of the range to be scanned.  For forward
	scans, it is in the form of the distance between the end of
	the range and Z;  for backward scans, it is the distance
	between the end of the range and B.  Thus, deleting
	or inserting characters never requires changing Q..2. 
	The use for changing Q..2 is moving the end of the range -
	for example, stopping the scan early (do ZU..2).
	Nothing forbids the macro to change point - that is a fine
	way to skip over characters that are "quoted" by others.
	  For efficiency, if the first of the 5 chars
	in the dispatch table is a space, the 5 are not macroed.
	Instead, the second character, minus 64, is added into
	Q..1, and the third is specially decoded.  " " means no
	action;  this feature makes to easy to skip over most
	chars, keeping track of horizontal position.  Other
	permissible third characters are "(" and ")".
	Their use is in counting parens or brackets.
	"(" means that if the scan is backwards and Q..1 is
	positive, the scan should terminate.  ")" means that if
	the scan is forward and Q..1 is negative, the scan should
	terminate.  If an open-paren-like character is given
	the dispatch " A(  " and the close is given " _)  ",
	the same dispatch table may be used to find the end
	of a balanced string going either forward or backward.
	"F^A" may be given 0 or 1 arg - it turns them into 2 the
	way "K", "T", etc. do.
<m>,<n>@F^A<q>
	the atsign modifier causes the scan to go backwards.
<m>,<n>:F^A<q>
	scans like F^A but expects its dispatch table to be a
	q-vector containing string pointers.  As with ordinary F^A,
	to each ASCII character there corresponds a word of the
	dispatch table, but this word should normally contain a
	pointer to a string to be executed when the character is
	seen.  Thus, you can macro strings longer than 5 characters
	without having to use up lots of q-registers to hold them.
	Alternatively, the dispatch table word may contain five
	characters starting with a space, which are interpreted as
	in an ordinary F^A.  Thus, simple characters that need only
	to be skipped over are no slower than in an ordinary F^A.
	The use of q-registers ..0, ..1, ..2 and ..3 is just like
	that in ordinary F^A.

F^B:
<ch>F^B<string>
	searches for the character <ch> in <string>.  <ch> should
	be the ASCII code for a character.  If that character
	does not occur in <string>, -1 will be returned.  If the
	char does occur, the value will be the position of its
	first occurrence (eg., 0 if it is the first char).

<pos>,<ch>F^<string>
	is like <ch>F^B but searches starting at position <pos>
	in the string.  If no occurrence is found past there, the
	value is -1.

@F^B<string>
	searches the buffer from point forward for a character
	NOT belonging to <string>, and then returns
	.,<address of that character>.  If such a character is not
	found, .,Z is returned.
-@F^B<string>
	searches the buffer backwards from point for a character
	NOT belonging to <string>, and then returns
	<address of that character>,<point>.  If no such character
	is found, B,. is returned.
<x>,<y>@F^B<string>
	searches the buffer from <x> to <y> (which can be either
	forward or backward) for a character NOT belonging to
	<string>, and then returns .,<address of character> or
	<address of character>,., whichever puts the smaller
	value first.
@:F^B<string>
	is like @F^B with no colon except that it searches for
	a character which IS a member of <string>.  It takes the
	same sorts of arguments that non-colon @F^B takes, and
	returns the same sorts of values.

F^E<string>
	replace <string> into the buffer at point.  Replacing
	means inserting, and deleting an equal number of
	characters so that the size of the buffer does not
	change.  The advantage of this command over
	"I<string> FKD" is that the gap need not be moved.
<n>F^E<string>
	replaces <string> in at <n>.  Point does not move.
	Like ".( <n>J F^E<string> )J".
<n>:F^E<q><string>
	replaces <string> into the string or buffer in q-reg <q>
	starting at the <n>'th character.  This is the only way
	that the actual contents of a string can be altered,
	although other commands copy pointers to strings, or
	create new ones.  If this command is done, it may be
	necessary to sweep the jump cache (see "F?") if the
	string being altered might be a macro that might
	contain "O" commands.

F^F:	F^F is the forward-only S-expression parser.
<s>,<e>F^F
	parses LISP s-expression from the starting value of point to
	address <e>, assuming that the state at the starting point is
	<s>, and returns the state at the ending point, leaving point
	there.

	A "state" includes all the essential information about
	the surroundings of a given spot in a LISP expression.  Its
	right half is the depth in parentheses.  Its left half is made
	up of several bits:
	  100,, => this spot is inside a comment.
	    4,, => this spot is between vertical bars.
	    2,, => the preceding character is an atom constituent or
		  slash, or is slashified, and we're not in a comment.
	    1,, => the following character is slashified.
	Usually it is fine just to use zero for all the bits.  Using a
	negative number as an initial state specifies a negative paren
	depth and makes all the bits zero.

	Several Q-registers are set up on return:
	  Q..0 receives the address of the last comment-start or
	    vertical-bar open encountered.  It is useful when you
	    find that the ending point is inside one or the other.
	  Q..1 receives the address of the last unmatched "(",
	    or -1 if there was none encountered.
	  Q..2 receives the address of the start of the last complete
	    s-expression (ie, its termination was passed), or -1 if no
	    s-expression begun has terminated.

	Scanning stops either at address <e>, or when the parenthesis
	depth becomes zero.  Aside from this termination condition,
	the depth is used only as a relative quantity, so you can
	choose the initial depth to obtain the desired termination
	condition, or make it far away from zero to disable the
	feature. 

	To use F^F to move forward, choose an initial parenthesis
	depth that will make it stop at the appropriate place.
	To use F^F to move backward,  go to the beginning of a
	top-level s-expression, or the beginning of the buffer, and
	scan forward to the point you are trying to move from.
	The value in Q..1 or Q..2 will tell you where to move to.
	To determine whether a line has a comment on it, scan from
	such a known point to the end of that line (before the CRLF).
	When the F^F returns, it will tell you whether that point is
	inside a comment.  If it is, Q..0 will address the semicolon.

:F^F	is like F^F, except that it stos scanning if it passes the
	start of an atom.  It also stops when plain F^F stops.
	It is useful for such things as making sure that an F^F
	doesn't run away from you because there was a very long
	s-expression where you expected an atom.

F^G:
<m>,<n>F^G<q>
	Performs a local variable value swap on an EMACS-format buffer
	table <q>.  <m> should be the index in <q> (in words) of the
	entry whose local variable values are to be swapped.  <n> is
	the offset within the entry of the first local variable.
	The number of local variables is determined from the length of
	the entry, which is the first word of the entry.
	Each local variable is described by two words:  the first is
	the name;  the second, the swapped-out value (the global one,
	for the selected buffer;  the local one, for other buffers).
	The name of a local variable is either a string, for a
	QFOO-type variable, or the :FSQPHOME of a short-name
	q-register or ^R-mode character definition or FS flag.

	Each swapped-out value is exhanged with the value cell of
	the variable, found by looking the name up in q-reg ..Q if it
	is a string, or by using it as an address if it is a number.
	A number must represent a ^R character, a q-register, or an FS
	flag.  Suitable numbers can be obtained by pushing the
	location involved and then looking at -1:FS QP HOME:
	[A -1:FS QP HOME( ]A) returns the address of q-register A;
	[^^@ -1:FSQP HOME( ]^^@), that of the definition of ^@;
	F[^R REPLACE -1FS QP HOME(F]^R REPLACE), the code for
	FS ^R REPLACE.  To turn such an address into a string
	containing the name of the q-register, use @FS QP HOME.

	The @ and : flags make F^G transfer data only onw way instead
	of swapping.  @F^G copies the current values into the local
	symbol table, but does not change the current values.  :F^G
	copies from the local symbol table setting the current values.

F^K<prompt>
	is a command that reads a string argument.  If the current
	macro was called by another macro, without the @ flag (see @),
	the string argument is read from the caller as if by ^]^X.  In
	this case, the <prompt> is ignored.  If the current macro was
	run as the definition of a ^R-mode character, or was called
	with the @ flag to pretend that it was called directly by ^R,
	the string argument is read by doing M*F Hook*.  The
	<prompt> string should be read as an argument by that macro,
	which the user is responsible for defining.  In addition, F^K
	can be given one or two numeric arguments which will be passed
	on to *F Hook*.
	In either case, the string argument is returned as a string
	object as the value of the F^K command.
	If you like, you can make the *F^K Hook* macro exit the
	macro containing the F^K in the event that the user rubs out
	past the beginning of his argument.  This can be done with
	-3 FS BACK RETURN.

:F^K	can be used by a macro to tell whether it was called from
	the internals of TECO or from another macro.  It returns -1
	if the macro was called from the internals of TECO or was
	called with @M.  Otherwise, it returns a positive number.

F^R	reports to ^R that completely unknown changes to the buffer
	have taken place, and that ^R should not assume that any of
	the information it remembers from previous display is still
	valid.
<m>F^R	is a no-op, to round out F^R with zero or two args, so that
	you can pass to F^R any args you might have returned to ^R,
	with equivalent results.
<m>,<n>F^R
	reports to ^R that the range of characters <m> to <n> has
	changes made in it, and needs to be redisplayed.  This is
	like doing <m>,<n>@V except that F^R never actually does
	the redisplay;  it just causes it to be done at the next
	opportuntity.  F^R should not be used unnecessarily.  For
	example, once a macro uses F^R, it cannot be used on any
	buffer except the one the next ^R up is displaying.

<m>,<n>@F^R
	reports to ^R that screen lines <m> through <n>, not
	including <n>, need redisplay.  This is used for exiting
	minibuffers.  It is very important that the text previously
	displayed on line <n> be unchanged in the buffer and still
	belong on that line.

:F^R	asks ^R to verify that the window in FS WINDOW is still
	valid, and choose a new valid one if it isn't.  This operation
	is extremely fast when the window is still valid, because it
	assumes that FS ^R VPOS is up to date.  If you want it to
	assume nothing, do a plain F^R first.  Normally, the previous
	window is kept if it is still valid.  To insist on
	re-centering point, do -1FS WINDOW first to flush the old
	window. 
<n>:F^R
	asks ^R to choose a new window putting point on line <n> of
	the window area.  <n> is relative to FS TOPLINE.  The old
	window is never re-used.  FS ^R VPOS is not depended on.
@:F^R	is like :F^R, except that on a
	terminal with insert/delete line capability ^R will attempt to
	move text on the screen IMMEDIATELY so as to optimize
	redisplay later.  The number of lines up or down to move the
	text is determined by assuming that FS ^R VPOS is valid.
<n>@:F^R
	is like <n>:F^R, but moves text on the screen if possible.

F^S:
<m>,<n>F^S<q>
	searches the buffer or word aligned string in <q> for a word
	containing <n>, starting at word <m>.  If one is found, its
	index is returned; otherwise, -1 is the value.  The index of
	one occurrence can be used as <m> in the next call to find the
	next occurrence.

<m>,<n>:F^S
	searches the ^R character definition table instead of a buffer.
	In other respects it is like F^S.

F^X	within a macro, this command returns as its values the
	arguments that were given to the macro.  As many values
	are returned as args were given.  To find out how many
	there were, use F^Y.

F^Y	returns a value saying how many args it was given.  For
	example, WF^Y returns 0;  W1F^Y, 1;  W1,F^Y, 2;  W1,2F^Y, 3.
	The : and @ flags are also indicated by bits in the value;
	:F^Y returns 4 and @F^Y returns 8.  1,2:@F^Y returns 15.
	The modifier flags, and the previous arguments, are flushed.

F^^<^R-char>
	F^^ takes the name of a ^R-character definition, as a
	q-reg, and returns the 9-bit character code for the
	character.  For example, F^^.^R. returns the code for
	Control-., which is 256 octal.

<string>:F^^
	:F^^ is used to test whether a string contains a valid short
	q-register name.  The string is passed as a string pointer in
	a prefix argument.  If the string contains a short q-register
	name then the :FS QP HOME of that q-register is returned.
	Otherwise, zero is returned.

F"<condition>
	F" is a conditional.  It works like ", except that
	whereas " throws away its argument after testing it,
	F" returns its argument, whether it succeeds or fails.
	Thus, QA-QBF"LW'+QB implements max(QA,QB).
	(a better way, though, is QA,QBF^@, before any command
	which ignores a pre-comma argument).

F$	is used to read or set the status of case conversion
	on input and case flagging on output, for terminals
	that do not have lower case.  What those features do
	when activated is described below.  F$ controls them thus:
	with no arg, returns the value of FS CASE 
	and inserts in the buffer before the pointer the
	case-shift char, if any, and the case-lock char, if any.
	With an arg, sets FS CASE  to that arg, and takes a string
	argument whose 1st char becomes the new case-shift,
	and whose 2nd char becomes the new case-lock.
	(if the chars are the same it is only a case-shift)
	(if there are no chars, you get no case-shift or -lock, etc)
	the old case-shift and case-lock, if any, become normal
	characters before the string arg is read.  Thus, repeating
	an F$ command will not screw up.

	Case conversion on input:

	When FS CASE is nonzero, all letters are normally
	converted to the standard case, which is upper case
	if FS CASE  is positive;  lower if negative.
	The case-shift char causes the next char to be read in
	the alternate case.  The case-lock char complements
	the standard case temporarily
	(it is reset for each cmd string).
	The case-shift quotes itself and the case-lock.
	The "upper case special characters" which are "@[\]^_"
	are not normally converted, but if one of them is preceded
	by a case-shift it will be case shifted into a "lower
	case special character" (one of "`{|}~<rubout>").
	(note that case conversion happens during command execution
	now, so that it makes sense to change modes in the middle
	of a command string.  However, no case conversion is done
	on characters that come from macros)
	(note also that it doesn't work well to have FS CASE
	and FS *RSET simultaneously nonzero, for complicated reasons).

	Case-flagging on output:

	If FS CASE  is odd, chars in the nonstandard case
	(and the "lower case special characters") will
	be preceded by case-shifts on typeout from the buffer.
	If FS CASE is even, no flagging is done.

F(	is like ( except that whereas ( returns no values,
	F( returns its arguments.  F( therefore facilitates
	putting the same information in two places without the
	use of a q-reg.
<n>F(<m>)
	converts <n> feet <m> inches to inches.

F)	resembles ), but whereas ) returns its arguments
	combined with the values stored by the matched (,
	F) returns precisely its arguments.  The data saved
	by the corresponding ( is discarded.

F*	reads and ignores a string argument.  Useful in macros
	because "F*^]^X" reads and ignores a string argument
	passed to the macro.

F+	clears the screen.  Like "^L", but does not separate
	pages in files.  If only a part of the screen is in
	use (FS LINES or FS TOP LINE is nonzero), only that
	part is cleared.  To be sure to clear the whole screen,
	bind both of those flags to 0 around the F+.  On
	non-erasable displays such as Tektronixes, this does clear
	the screen, even though TECO does not otherwise treat such
	terminals as displays.  On printing terminals, a CRLF is
	typed.

F6<string>
	returns a word of SIXBIT containing the first six
	characters of <string>.
<sixbit>F6
	interpreting <sixbit> as a word of SIXBIT, converts
	it to ASCII which is then inserted in the buffer
	before the pointer.
<sixbit>: F6
	returns a string containing the characters of <sixbit>.
	@F6 is an obsolete equivalent command.

F;<tag>
	is a "throw", a la LISP.  See "F<" below.

F<!<tag>! ... >
	is a catch.  If anywhere in the arbitrary code which
	may replace the "..." a "F;<tag>" command is executed,
	control will transfer to after the ">" that ends the
	catch.  If no "F;" is executed, the catch acts like
	an iteration, so if the code should be executed only once,
	"1F<" should be used.  When a "F;" or throw happens,
	all macros, iterations and errsets
	entered within the catch are exited and the
	q-reg pdl is unwound to the level it had at the time
	the catch was entered.  Example:
	    F<!FOO! [A FIUA QAI QA-32"E F;FOO' ]A>
	reads characters from the terminal and inserts them, up to
	but not including the first space, and does not modify
	the q-reg pdl (never mind that this macro might be improved).
	If a throw ("F;") is done to a tag that does not belong
	to any catch containing it, an error "UCT" occrurs,
	at which time nothing has been unwound.
	The ">" ending a catch will return 0 if the catch was
	exited normally;  if it was thrown out of, the argument
	given to the throw will be returned (or 1 if there
	was no argument).	
	Note that case is not significant in the F; or in the F<.
:F<	is an errset and a catch at the same time!
	Amazing what happens when your program works
	by simply examining a bunch of flags!

F=	does an ordered comparison of strings.
	If "F=" has numeric args, they specify the range of buffer
	to be used as the first comparison string.  Otherwise,
	the "=" should be followed by the name of a q-reg which
	should hold the first comparison string.
	The second comparison string should follow the command as
	a string argument, as for the "I" command.  (the @ modifier
	works just as it does for the "I" command)
	the two strings are compared, and if they are equal
	0 is returned as the value of the "F=" command.
	If the first string is greater, a positive value
	is returned;  if the second, a negative value.
	If the value isn't 0, its absolute value is 1 +
	the position in the string of the first difference
	(1 if the first characters differ, etc.).
	A string is considered to be
	greater than any of its initial segments.


F?	mbox control;  argument is bit-decoded.
	No arg, or arg=0, implies arg=30 .

	bit 1.1 - close gap.
		May be needed for communication with other
		programs that don't understand the gap.

	bit 1.2 - GC string space.
		Useful before dumping out, or if it is suspected
		many strings have recently been discarded.

	bit 1.3 - sweep the jump cache.
		Necessary if a string's contents have been
		altered by the F^E command, and might be a macro
		that might have contained "O" commands.
		Also necessary if :EJ is used after increasing
		the value of FS :EJPAGE (thus replacing one
		file with another in core).

	bit 1.4 - flush unoccupied core.
		Good to do every so often,
		or if it is likely the buffer has just shrunk.

	bit 1.5 - close the gap, if it is > 5000 characters long.
		It is good to do this every so often,
		in case the user deletes large amounts of text;
		say, whenever excess core is flushed.

FA	performs text justification on a range of the buffer
	specified by 1 or 2 args (as for K, T, commands, etc.).
	The idea is that whenever you edit a paragraph,
	you use FA or a macro that uses FA to re-justify it.
	The line size is kept in FS ADLINE .
	A CRLF followed by a CRLF, space or tab causes a break.
	So does a CRLF, space or tab as the first character
	of the text being justified.
	An invisible break can be produced before or after a
	line by	beginning or ending it with space-backspace.
	CRLFs that do not cause breaks are turned into spaces.
	Excess spaces that wind up at the end of a line are deleted.
	Other excess spaces are retained.  Thus, to win, you should
	insert spaces after all ends of sentences which are at the
	ends of lines, before you call FA.  You might insert spaces
	at the ends of all sentences, but you can presume that they
	are already there, when desired, in the middle of the line.
	Spaces at the beginning of a line are treated as part of
	the first word of the line for justification purposes,
	to prevent indentation of paragraphs from changing.
	The last part-line of stuff to be justified is only filled.
	Tabs prevent justification of what precedes them on a line;
	to prevent filling, you must put in a break (use sp-bs).
	I suggest using "FA" in the following macro:
	    [0 Z-^YU0
	        ^XJ <.,Z-Q0FB.
	^O?
	;  :0L I >
	    !""Make sure .'s and ?'s at end of line have spaces!
	        ^X,Q0FA	!actually justify!
	    ]0
	If you want indented paragraphs, simply indent them the
	right amount when you type them in.  "FA" will leave the
	indentation alone.  "FA" knows about backspace.
	Sometimes it is desirable to put a space in a word.  To
	do that, use space-backspace-space.
	Tab characters in the buffer or in typeout are counted
	using tab stops at a spacing controlled by FS Tab Width.
@FA	like "FA" but only fills (doesn't justify)

FB	bounded search.  Takes numeric args like K, T, etc.
	Specifying area of buffer to search, and a string 
	argument like S, N, etc.  The colon and atsign flags
	are used as they are by other search commands.
	:FB is like :S, not like :L.  That is, :FB returns a
	value indicating extent of success, and searches the
	same range of the buffer as FB with no :.
	If two args in decreasing order are given searching is
	done in reverse.  With one negative arg, the search is
	forward, but through a range that ends at the pointer.

FC	takes arguments like K and
	converts the specified portion of the buffer to
	lower-case.  Only letters are converted.
@FC	converts a specified portion of the buffer to upper case.

<ch>:FC	returns the upper-case equivalent of the character whose
	numeric code is <ch>, as a numeric code.  Meta-bits in
	<ch> are passed through unchanged.
	:FC may also be given a string pointer as argument.
	It returns a new string containing the upper-case of its
	argument.

FD	a list manipulating command whose main use is in
	"<arg>FDL", which moves down <arg> levels of parens.
	FD returns a pair of args for the next command.
	If <arg> is positive, they specify the range of the
	buffer from the pointer rightward to the first
	character that is <arg> levels up;
	if negative, leftward to the first character
	-<arg> levels up.

FE	inserts a list of TECO error messages and explanations
	in the buffer before the pointer, one message per line.

<arg>FE	inserts only the line describing the error
	of which <arg> is the error code.
	<arg> might have been returned by an errset,
	or might be the value of FS ERROR .
	Since error codes are actually strings, <arg>FE
	is equivalent to G(<arg>) I<crlf>

@FE<errname>
	returns the error code associated with the given
	error name <errname>.  Only the first three characters
	of <errname> are used.  This is useful for analysing
	anticipated possible errors and recovering appropriately.
	Another way to do that is to compare the first three
	characters of the error code, which is a string, against
	the expected ones, with F~.

:FE	inserts a list of FS flag names in the buffer before
	the pointer, one name per line.

FG	does error processing.  With no argument, it simply
	rings the terminal's bell.  Given the @ modifier, it
	also throws away type-ahead.  Given a nonzero numeric
	argument (which should be s string), FG prints its
	contents as an error message (obeying FS VERBOSE).
	Accompanying the numeric argument with the : modifier
	causes the error message to be typed at the top of
	the screen (think of : FT).  Unlike most commands that
	do typeout, FG does not change Q..H, so that typing an
	error message will not inhibit buffer display or ^R
	redisplay.  Instead, FG sets FS ERRFLG so that the
	next buffer display will not overwrite the line(s)
	occupied by the error message.

FI	input one character from the terminal and return its
	ASCII value.  (same as vw without the v)
	if the "mode" (in q-reg ..J) has changed, the new
	value will be put on the screen, unless input is
	already waiting when the FI is executed.
:FI	similar to FI, but doesn't flush the character.
	It will be re-read by the next FI
	or by TECO's command string reader.
@FI	is like FI, but returns a character in the 9-bit
	TV character set, rather than converting it to ASCII
	as FI does.
	In the TV character set, the 400 bit means "meta",
	the 200 bit means "control", and the bottom 7 bits
	are a printing character (if < 40, it is one of the
	new TV graphics, or else it is a formatting character).
	Note that there exist controlified lower case letters
	different from their upper case counterpart (for example,
	341 octal is control lower case a).
@:FI	analogous.

FJ	insert the TECO job's command string as read from DDT
	in the buffer.  Will normally end with a CR-LF
	but may be null.

FK	returns minus the value of FS INSLEN;  that is,
	minus the length of the last string inserted by "I", "G" or
	"\", or found by a search or "FW".  FK is negative except
	after a successful backward search, or backward "FW".
	Thus, "SFOOFKC" will move to the beginning of the FOO found.
	"-SFOOFKC" will move to the end of the FOO found.
	See also the ^B command in this context.
	"SFOOFKDIBAR" will replace FOO with BAR.  See "^F".
	"IBLETCHFKC" inserts BLETCH and backs over it.

FL	parses lists or S-expressions:
<arg>FL	a list maniulating command, that returns 2 values specifying
	a range of the buffer.  If <arg> is >0, the range
	returned is that containing the next <arg> lists
	to the right of the pointer;  if <arg> is <0,
	the range is that containing <arg> lists
	to the left of the pointer.  This command should be
	followed by a command such as K, T, X, FX ...
	which can take 2 args;  the specified number of lists
	will be deleted, typed, put in q-reg, etc.
	To move to the other side of the lists, do "<arg>FLL".
	The syntax parsed by FL is controlled by the delimiter
	dispatch table in Q..D;  the character types known
	are "A", " ", "|", "/", "(", ")" and "'", and any character
	can be redefined to be of any of those types.

:FL	is like FL, but stops before the ( that starts the list
	instead of after.  -:FL stops after the ), etc.

<arg> @FL
	is like "<arg>FL", but refers to <arg> s-expressions
	rather than <arg> lists.  An s-expression is either
	a list or a LISP atom, whichever is encountered first.
	"@FW" is used to find LISP atoms when necessary.
	@:FL stops before the next s-expression - if it is a list,
	it stops before the (, and if it is quoted it stops before
	the quote.

FM:
<m>,<n>FM
	attempts to move the pointer so that the cursor will
	appear at hpos <n>, <m> lines below where it started out.
	"FM" without the "@" modifier can move only toward the
	end of the buffer.  It operates by moving the pointer
	downward in the buffer until either 1) the exact desired
	absolute hpos and relative vpos have been reached, in
	which case "FM" simply returns, or 2) the end of the buffer
	is reached, which causes a "NIB" error, or 3) the line
	below the desired one is reached, in which case it is
	known that the desired combination of hpos and vpos does
	not exist, so FM reverses its motion until it is back
	on the desired line, then issues a "NHP" error.
	"FM" tries to avoid leaving the cursor between
	a CR and the following linefeed.
	"FM" will not currently work if ^R mode has never been
	entered, but it need not be in ^R mode.
	The ":" modifier causes "FM" to accept any hpos greater
	than or equal to the second argument as a condition for
	success, rather than demanding exact equality.
	The "@" modifier causes "FM" to scan toward the beginning
	of the buffer rather than toward the end.  The first argument
	should not be positive.  The algorithm
	is otherwise unchanged and ":" has the same meaning
	(accept any hpos >= the specified one).

FN	is the same as [..N :I..N.  It is needed because it
	elminates the possibility of a ^G-quit between the
	push and the insert.  If such a quit happened, the
	previously set up undo action would be performed
	twice instead of once, and that might have bad results.
	To perform the opposite action - pop and macro Q..N -
	just do "-FS QPUN".
	The ..N macro has no effect on the value(s)
	returned by the macro that set it up.
	The atsign flag allows the user to specify a string
	delimiter, as with the I command.

FO<q><name>
	binary-searches tables of fixed-length entries.
	It is intended for searching and constructing symbol tables.
	<q> should be a q-vector or pure string containing
	the table, and <name> the item to search for.
	The table's data must be an integral number of words.
	The first word of the table must contain the number of words
	per table entry;  the rest of the table is then divided into
	entries of that size.  TECO initialized this size to 2 words
	per entry, but it can be made larger.  The first word of each
	entry should be the entry's name, as a TECO string pointer.
	This name is what
	FO will match against its string argument.  The second word
	of each entry should be the value;  the use of any extra
	words is up to the user.  The entries' names must always be
	kept in increasing order, as F~ would say, or FO's binary
	search will lose.  Also, they should not contain leading,
	trailing, or multiple spaces, or any tabs.  Their case is
	ignored.
	  FO, without colon, will return the value from the entry if
	the name is found;  otherwise, an UVN or AVN error results.
	  @FO allows only exact matches.  Abbreviation is not allowed.
	  :FO returns the offset (in words) of the entry found;
	if the name is not found, :FO returns minus the
	offset (in words) of the place the name ought to be inserted
	in the table.  The offset of the first entry in the table is
	1, to skip the word in the front that contains the entry
	size.
	  <arg>FO is like plain FO except that if the name is
	undefined <arg> is returned as its "default value".
	Ambiguous names still cause errors.

	  Here is a macro that uses FO to create variables that
	can then be used with the Q<name> construction:
		[0 :I0		! Get variable name in Q0 !
	   ! Find it, or where to put it if not found !
		[1 :FO..Q0U1
		Q1"L			! If not found, put it in !
		   Q..Q[..O		! Symbol table lives in ..Q !
		   -Q1*5J 10,0I 10R	! Make space at right place !
	   ! Install string containing variable name !
		   Q0,.FSWORD'
	If this macro is put in QV, then MVFoo will create
	a variable named Foo.
	  If the table is a pure string, the data must start on a word
	boundary, which means that the string's header must start in
	the second character in its word.  In addition, the pointers
	to the entries' names are taken to be relative to the table
	itself.  That is, the "pointer" should be an integer which,
	when added to the TECO string pointer to the table, should
	give a TECO string pointer to the name of the entry.

FP:
<obj>FP	returns a number describing the data type of the
	object <obj>.  The possible values are (in DECIMAL):
	  -4	A number, not even in range to point into
		 pure or impure string space.
	  -3	A number that is in range for pure string space
		 but does not point at a valid string header.
	  -2	A number that is in range for impure string space
		 but does not point at a valid string header.
	  -1	A dead buffer.
	  0	A living buffer.
	  1	A Q-vector.
	  100	A pure string.
	  101	An impure string.

FQ<q>	returns the number of characters in q-reg <q>
	(-1 is returned if the q-register contains a number).

FR	tells TECO to update the displayed mode from q-reg ..J,
	provided it has changed, and no terminal input is available.
	:FR erases the mode line.


FS	reads in a flag name as a text argument.
	Flag names may be any length, but only the
	first six characters are significant.
	Spaces are totally ignored.  Only enough
	of the flag name to make it non-ambiguous is required.
	However, in programs, abbreviation should be minimized.
	The result of the FS is the current value of the flag.
	If an argument is given to the FS and the flag can be set,
	is then set to that value.
	If a flag can be set, to make it's value the second operand
	of an arithmetic operator put the FS command in parens.
	Otherwise, FS will think it has an arg and set the flag.
	Flags labeled "read-only" do not require that precaution.
	Flags currently implemented are:

FS % BOTTOM	specifies the size of the bottom margin as a
		percentage of the number of lines being displayed.
		Initially 10.  Rather than let the cursor appear
		inside the bottom margin, TECO will choose a
		new buffer window - unless the bottom of the
		buffer appears on the screen already.

FS % CENTER	specifies where TECO should prefer to put the
		cursor when choosing a new window,  as a
		percentage of the screen from the top.
		Applies even if the end of the buffer appears
		on the screen - in fact, the purpose of this
		variable is to make sure that when you go to
		the end of the buffer some blank space is provided
		to insert into without total redisplay.
		Initialy 40.

FS % END	specifies (as a percentage of total size) the
		size of the area at the bottom such that TECO
		should never choose a new window putting the
		cursor in that area.  Initially 30.

FS %OPLSP	(read only) nonzero if the terminal is inferior
		to a LISP job (actually, if the superior has set the
		%OPLSP bit in the .OPTION variable of the TECO).

FS % TOP	the size of the top margin.  Analogous to
		FS %BOTTOM.  Initially 10.

FS %TOCID	(read only) nonzero if the terminal can insert
		and delete characters.  This flag reflects
		the bit with the same name in the terminal's
		TTYOPT variable, and it is updated whenever
		TECO is restarted or FS TTY INIT is done.
FS %TOFCI	(read only) nonzero if the terminal can generate
		the full 9-bit character set.  See FS %TOCID.
FS %TOHDX	(read only) nonzero if the terminal is half-duplex.
		See FS %TOCID.
FS %TOLID	(read only) nonzero if the terminal can insert
		and delete lines.  See FS %TOCID.
FS %TOLWR	(read only) nonzero if the terminal can generate
		lower case characters.  See FS %TOCID.
FS %TOMOR	(read only) nonzero if the user wants --MORE--
		processing, in general.  See FS %TOCID.
FS %TOOVR	(read only) nonzero if the terminal is capable of
		overprinting.  See FS %TOCID.
FS %TOROL	(read only) nonzero if the user has selected scroll
		mode.  See FS %TOCID.
FS %TOSAI	(read only) nonzero if the terminal can print
		the SAIL character set.  See FS %TOCID.

FS *RSET	initially 0.  Nonzero suppresses automatic
		unwinding of TECO's various pdls each time
		through the top level loop.  In other words,
		when FS *RSET is non-zero, errors not caught by
		errsets enter break-loops in which q-regs may
		be examined (unless the user's error-handler
		macro in q-reg ..P intervenes).  The break-loop
		may be returned from with ")^\", or thrown
		out of with "^W" or "F;".  The suspended program
		and its callers may be examined with FS BACKTRACE.
		For more info on break-loops, see under q-reg ..P.

FS .CLRMOD	(normally -1) if < 0, TECO clears the screen
		whenever it gets the terminal back from its superior.
		If 0, that is not done
		(used mainly for debugging TECO).
		If > 0, screen clearing is totally eliminated,
		even if requested by the program
		(use this for debugging macros that try to
		destroy trace information).

FS .KILMOD	normally -1.  If 0, FS BKILL doesn't actually kill.

FS .TYI BACK	backs up the pointer FS .TYI PT by one step.  Back up
		the pointer n steps and then you can use FS .TYI NXT
		n times to get the n previous input characters.

FS .TYI NXT	is used to extract from the ring buffer of saved
		type-in characters.  See FS .TYI PT.

FS .TYI PT	is TECO's pointer into the ring buffer of the last
		60 input characters.  All input characters read by
		TECO are saved in this buffer, for the sake of
		programs to tell the user what he typed by mistake.
		This pointer is used in conjunction with FS .TYI BACK
		and FS .TYI NXT, for extracting the contents of the
		ring buffer.  This is done with the following code:
		    FS .TYIPT[1 [2
		    < FS .TYINXT U2
		      do something to print the character in Q2.
		      FS .TYIPT-Q1@;>
		Each FS .TYI NXT extracts the next character from
		the buffer and advances the pointer.  When the pointer
		has been advanced all the way around to its original
		position, the entire buffer has been printed.

		After an input character is read and stored in the
		buffer, FS .TYI PT is initialized to point at that
		character.  Though FS .TYI PT is changed by ths use
		of FS .TYI NXT and FS .TYI BACK, this has no
		effect on how input characters are stored.

FS :EJ PAGE	is the number of the lowest page used by :EJ'd
		shared pure files.  Initially 256.  If multlipied
		by 5*1024, and then added to 400000000000 (octal),
		the result is a suitable string pointer to the
		last file :EJ'd.  :EJ looks at this flag to
		figure out where to insert the next file to avoid
		clobbering the previous ones.  It is an error to set
		this flag explicitly to a value lower than it already
		has;  when it is set to a higher value, the
		intervening pages are removed from the address space.

FS :ET MODE	(Twenex only) controls how filename defaults are used
		by the :ET command.  A bit mask of version, type, name,
		directory, and device (from 1 to 20 respectively).

FS ADLINE	is the line-size used by the FA command.

FS ALTCOUNT	the number of 's that TECO has seen at
		interrupt level.
		That is, an approximation to the number of command
		strings that the user has typed ahead.
		Useful in user-defined buffer display macros
		(q-reg ..B).

FS BACK ARGS	(read-only) returns the arguments of a macro call
		level up on the stack.  It returns 0, 1, or 2 values,
		as F^X does.  The argument to FS BACK ARGS is
		either the positive number of a frame counting from
		the bottom (0 = the outermost frame), or a negative
		number counting back from the currently executing
		macro (which, by the way, has no stack frame and
		cannot be referred to in either way).

FS BACK DEPTH	(read only) is the number of macro call frames on the
		stack.  This does not include the currently executing
		macro.

FS BACK PC	returns the PC of a macro call frame.  Which frame,
		is specified as in FS BACK ARGS.  The PC of a frame
		is the number of the next character to be executed,
		relative to the beginning of the string.

<m>,<n>FS BACK PC
	        sets the PC of the frame specified by <n>, to <m>.

FS BACK QP PTR	(read-only) gives the level of the bottom of a
		macro call frame's q-register pdl frame.  This is the
		level which a ^\ done in that macro will unwind to.
		The macro call frame is specified as for FS BACK ARG.

FS BACK RETURN	(write-only) returns control to the macro call frame
		specified as for FS BACK ARG.  If control left that
		frame not with an M but via an exit from within a TECO
		command (such as a ^R), control is returned to that
		command.  -1 FS BACK RETURN is equivalent to ^\.

FS BACK STRING	(read-only) returns a pointer to the string or buffer
		being executed by macro frame specified as for
		FS BACK ARG.  Sometimes a macro frame will be
		executing a string constant internal to TECO.  There
		cannot be a string pointer to such an object, so a
		byte pointer to the object (a number, according to FQ)
		will be returned instead.  All you can do with such
		values is compare them.

FS BACKTRACE
		used to see what program is running at higher
		levels of the mcro pdl.  The program is inserted
		in the buffer, and point is put at the place
		it is executing.  The macro call frame is specified
		just as for FS BACK ARGS.

FS B BIND	is useless, but F[ B BIND and F] B BIND are useful
		for pushing to a temporary buffer, or popping back
		from one.  F[ B BIND pushes ..O, then does FS B CREATE,
		but with the extra feature that if an error happens
		instead of just popping back ..O, the temporary buffer
		will be killed.  That is because instead of doing ]..O,
		F] B BIND will be done, which is just like FS B KILL
		with no argument.
		If after creating a buffer with F[ B BIND you change
		your mind and want to keep it, pop the previously
		selected buffer off the pdl with the ] command.  The
		F[ B BIND will no longer be on the stack to kill the
		new buffer when you return.
		An argument to F[ B BIND specifies the initial size
		of the termporary buffer in characters.

<n>FS B CONS	returns a newly cons'ed up buffer <n> characters long.
		The contents are all initially zeros, and the pointer
		starts out at the beginning of the buffer.  If <n> is
		not specified, 0 is assumed.
		When a buffer is newly created it is at the
		top of memory.  The closer a buffer is to the
		top of memory, the more efficient it is to do
		large amounts of insertion in it.

FS BCREATE	is like FS B CONS U..O - the buffer is selected
		instead of returned.

FS BKILL	see q-reg ..O.  FS BKILL is used for freeing
		buffers explicitly.  With an argument, it frees
		the argument, which should be the result of
		applying the q command to a q-reg containing a
		buffer.  Attempting to kill the currently
		selected buffer is an error.  For example,
		QAFS BKILL kills the buffer in qa.  After that is
		done, qa still contains a buffer pointer, but it
		has been marked "dead".  If there were other pointers
		to the same buffer in other q-regs, TECO will regard
		them too as "dead" buffer pointers.  An attempt to
		select the buffer using one of those pointers
		will result in an error.
		FS BKILL may be used without an
		argument, in which case it pops the q-reg pdl
		into Q..O, and if the new value of Q..O is
		different from the old, the old value is
		killed.

FS BOTHCASE	(initially 0) >0 => searches ignore case of letters.
		That is, the case used in the search string is
		irrelevant, and either lower or upper case will
		be found.
		<0 => searches ignore case of special characters
		also ("@[\]^_" = "`{|}~<rubout>").

FS BOUNDARIES	reads or sets the virtual buffer boundaries
		(this command returns a pair of values)
		the virtual boundaries determine the portion of
		the buffer that most other commands are allowed
		to notice.  Normally the virtual boundaries
		contain the whole buffer.  See the B, Z and H
		commands, and FS V B and FS V Z.

FS BS NO LF	if nonzero inhibits the LF that follows backward
		motion or rubbing out, in ^R mode on printing terminals.

FS CASE	like F$ (F-dollar) but neither inserts the case-shift
		and case-lock if no arg, nor expects a string arg.
		That is, it gets or sets the numeric quantity which
		determines the standard case and whether to flag
		on output.
		
FS CCL FNAME	(Twenex only) a string, in the same format as FS D FILE,
		of the jfn given in AC1 if TECO was started at the CCL
		entry point; or zero if it was not or the filename has
		already been read.  Discards the JFN unless given an
		atsign modifier.

FS CLK INTERVAL
		is the interval between real-time clock
		interrupts, in 60'ths of a second.  If it is zero,
		real-time interrupts do not happen.  Note that
		real-time interrupts are not actually processed
		unless/until TECO is waiting for input.  Their main
		use is for saving the buffer if the user walks away
		from the terminal without saving it.  FS IN COUNT may
		be useful with real-time interupts, and FS MODIFIED.
		Setting FS CLK INTERVAL postpones the next clock
		interrupt to one full interval in the future.

FS CLK MACRO   is the real-time interrupt handler macro.
		If the macro types out, it must not leave Q..H set.

FS CTL MTA	if negative suppresses the ^R-mode definitions of
		all control-meta-letters (and ctl-meta-[, \, ], ^
		and _) to make it easy to insert control characters.
		This mode is convenient for editing TECO commands.

FS DATA SWITCHES
		(read only) the contents of the PDP10 console switches.

FS DATE	(read only) is the current date and time, as a number
		in file-date format.  It can be fed to FS FD CONVERT
		or to FS IF CDATE.

FS D DEVICE	is the current default device name, as a numeric
		sixbit word.  See F6.

FS DD FAST	(read only) is nonzero if the current default device
		is a fast device - that is, the local machine's disk.

FS D FILE	is the current default filename (that ER would use)
		as a string pointer.  Do G(FS D FILE) to insert it in
		the buffer.  This flag is very useful for pushing and
		popping with F[ and F].  The exact format of the string
		is the SNAME, a semicolon, a space, the device name,
		a colon, a space, the FN1, a space, and the FN2.
		To extract specific names from the string, search for
		spaces (or colon or semicolon) not preceded by a ^Q.
		Note:  on ITS, the default device name will never be
		DSK:.  It will be the name of the local machine
		instead.  This is so that comparison of two filenames
		which have been passed through the defaults and read
		out with FS D FILE will not say they are unequal
		because one has device DSK: and the other the machine
		name.

FS D FN1	is the current default first file name, as a numeric
		sixbit word.  See F6.

FS D FN2	is the current default second filename, as a numeric
		sixbit word.  See F6.

FS D FORCE	is to be used to "force out" a buffer redisplay for
		which the normal mechanism of deferring redisplay for
		input and later resumption will not do the right thing
		(such as, because the buffer being displayed is not
		the one permanently selected).  Setting FS D FORCE
		nonzero has two effects:  it forces the redisplay
		through to completion without considering the presence
		of input, and it inhibits updating the mode line.

FS D SNAME	is the current default SNAME, as a numeric sixbit
		word.  See F6.

FS D VERSION	is the current default file version number;  actually,
		a reflection of FS D FN2.  If the default FN2 is
		numeric, FS D VERSION is that fn2 as a number.
		If the default fn2 is ">" or "<", FS D VERSION is
		respectively 0 or -2.  Otherwise, FS D VERSION is -1.
		Writing -1 into FS D VERSION has no effect; writing
		any other value sets the default fn2 appropriately.
		On Twenex, versions are always numeric, so all values
		have their usual meaning.

FS D WAIT	if set nonzero causes TECO to pause slightly before
		each line of output.  This should give better overall
		response on slow displays.  It will make things much
		worse on fast displays, and slightly worse on printing
		terminals.  This flag is initialized by FS TTY INIT
		and at startup, according to the terminal speed.

FS ECHO ACTIVE	if nonzero says that output has been printed in the
		echo area during this ^R command, so ^R should clear
		the echo area when it next wants to read a command.
		Clearing is not done by the FI command;  only when
		the next distinct ^R command is to be read.
		Commands which wish to type in the echo area and not
		have their output erased can set this flag to zero.
		The FG command does, when it types in the echo area.
		FS ECHO FLUSH must be nonzero to enable this feature.
		Otherwise, the echo area is never cleared gratuitously.

FS ECHO CHAR	when a ^R command starts being executed, holds the
		character that invoked it;  set to -1 if anything is
		typed out;  when the command returns to ^R, if still
		not -1, it is echoed.  The command can set this flag
		itself to control the echoing.
		A string may be put into this flag by a macro.
		Then, if the command must be echoed, the whole
		contents of the string will be typed.

FS ECHO DISPLAY
		(write only) like FS ECHO OUT, but outputs in display
		mode, so that ITS ^P-codes may be sent.  See
		.INFO.;ITS TTY for details on available options.

FS ECHO ERRORS	if nonzero, causes error messages to be printed in the
		echo area.

FS ECHO FLUSH	if nonzero, enables the automatic clearing of the echo
		area after each ^R command which uses it.  See
		FS ECHO ACTIVE for details of how to control this.

FS ECHO LINES	the number of lines at the screen bottom to be
		used for command echoing.  Default is 1/6
		of screen size, except 0 on printing terminals.  If
		this flag is set to -<n>, echoing is turned off,
		and there are <n>-1 echo lines.  Thus,
		 -1FS ECHOLINES makes no echo and no echo area;
		 -5FS ECHOLINES makes no echo but a 4 line echo area;
		 0FS ECHOLINES makes echo but no echo area.
		Even if echoing is off, FS ECHOOUT may be used.

FS ECHO OUT	(write only) is used for outputting to the echo area.
		If the argument is a number, it is taken as the ASCII
		code for a character to be typed.  If the argument is
		a string, all the characters in it are typed.
		Characters are output as they would actually echo.
		Thus, sending a CR will actually do a CRLF,
		and sending a ^A will print either downarrow
		(in :TCTYP SAIL mode) or caret-A.

FS ERR		same as FS ERROR if read;  if written, causes an
		error with the error code that is written in it.
		Thus, to cause a "You Lose" error with 3-letter
		code LUZ, do :I*LUZ<tab>You Lose  FS ERR.  Your
		error message should not contain any CRLFs.
		Users who wish to generate errors themselves with
		the same codes that TECO uses should use TECO's
		standard strings for those errors (that is,
		do  @FE IFN FS ERR) so that comparing FS ERROR
		against @FE values will still work.

FS ERRFLG	is used to signal the buffer display routine
		(whether built-in or user-written) that an error
		message is on the screen and should not be overwritten.
		Its value is -<n> if the first <n> lines contain
		an error message, or nonnegative if there is none.
		If typeout is done when FS ERRFLG is negative, TECO
		will not actually type the first <n> lines of it.
		The <n>+1st line of typeout will appear in its
		normal position, beneath the error message.
		By that time, FS ERRFLG will be zero again.

FS ERROR	the error code of the most recent error.
		Errors caught by errsets are included.
		A TECO "error code" is now just a string
		containing the text of the error message. 
		Everything up to the first tab is the "brief" part
		of the error message;  if FS VERBOSE is 0, that
		is all that TECO will print out.  You can see if an
		error was (for example) an "IFN" error by doing
		F=(FS ERROR)IFN  and seeing if the
		result's absolute value is 4.  It still works to
		compare against @FEIFN, which returns the
		standard string that TECO always uses for
		internally-generated IFN errors.
		  I.T.S. I/O errors now have
		messages starting with "OPNnnn" where nnn is the
		I.T.S. open-failure code.  Macros which used to
		decode I/O errors by numeric comparison must
		switch to using F=, since the strings for such
		errors are consed up by TECO as needed. 

FS ERR THROW	(write only) return control to the innermost
		error-containing command loop.  This could be either
		an error catch command (:@< ... >), a ^R level, or the
		top level TECO command loop.  This command is the
		correct way for a user error handler (..P) to abort
		the computation which got the error.  The argument
		given to FS ERR THROW is returned from the error
		catch, if it happens to throw to an error catch.
		Otherwise, the argument is discarded.

FS EXIT	(write only) does a .BREAK 16, using the argument
		to FS EXIT as the address field.  See DDT DOC
		for a description of what the arguments mean.
		"100000." is a good value.  AC 2 will contain
		the address of the 7-word "buffer block"
		describing the current buffer.
		See the section "buffer block" at the end.

FS EXPUNGE	(T(w)enex only) expunges the connected directory.
		FS MSNAME can be used to connect to a directory so as
		to expunge it.

FS FDCONVERT	converts numeric file dates to text and vice versa.
		If there is a numeric arg, it is assumed to be
		in the format for its file dates, and converted
		to a text string which is inserted in the buffer.
		The form of the string is dd/mm/yy hh:mm:ss.
		In this case, no numeric value is returned.
		If there is no arg, a text string is read from
		the buffer starting at ., and . is moved over
		the string.  The string should be in the format
		inserted by FS FDCONV with argument, and will be
		converted to a numeric file date which will be the
		value of FS FDCONV.  See FS IFCDATE and FS OFCDATE.
		<n>:FS FD CONV returns a string containing the
		printed representation of the file date <n>, instead
		of inserting it in the buffer.  Twenex uses ODTIM and
		IDTIM, and so is less picky about date formats it will
		accept.  Also, a second argument to FS FSCONV is used
		as the ODTIM format flags.

FS FILE PAD	the character used to pad the last word of files
		written by TECO.  Normally 3 (for ^C).

FS FLUSHED	is nonzero if a --MORE-- has been flushed, and
		type-out is therefore suppressed.  The flag is
		positive if the flushage was due to a rubout,
		negative otherwise.  You can stop generating
		type-out when FS FLUSHED is nonzero, or you can
		clear it to make type-out start actually appearing
		again.

FS FNAM SYNTAX	controls TECO's filename readed.  If 0, when only
		one filename is present, it is used as the fn2
		(this is the default).  If positive, a lone filename
		is used as the fn1.  If negative, a lone filename is
		used as the fn1 and automatically defaults the fn2
		to ">".  The default TECO init file
		uses this flag to process a DDT command line.

FS FORK JCL	(Twenex only) gives a command line (jcl) for the inferior
		fork invoked via the next FZ command.

FS GAP LENGTH	(read only) the length of the gap.
		This is the value of EXTRAC (see "buffer block").

FS GAP LOCATION (read only) the buffer position of the gap.
		This is GPT-BEG (see "buffer block").

FS HEIGHT	(read only) number of lines on the screen, on display
		terminals (including --MORE-- and command lines).
		On printing terminals, wil be a very large number.

FS HELP CHAR
	contains the character to be used as the HELP character.
	Normally, it contains the code for C-_.
	However, you may want to bind it to -1 from time to time
	to prevent HELP from being recognized
	(such as, in the C-Q command).

FS HELP MAC	is a macro to be executed if the HELP character is
		typed.  HELP is Top-H on TV's, ^_H on any terminal.
		If FS HELP MAC is zero, the HELP character is
		ignored as a ^R command, and returned (as 4110 octal)
		by FI and VW.

FS H POSITION	(read-only) returns the number of character positions
		there would be to the left of the type ball
		if the contents of the buffer (or at least
		everything after the previous carret) were
		printed on a hardcopy terminal with backspaces really
		backing up and tab stops every FS Tab Width columns.

FS HSNAME	is the user's home directory.  The home directory
		is a little more permanent than the working directory
		(FS MSNAME) and is used for storing things that
		permanently belong to the user, such as the RMAIL
		file, which do not have their own special directories
		the way init files do.

FS I&D CHR	if nonzero, tells TECO to try to use the insert and
		delete character operations to speed redisplay.
		Initialized nonzero if the terminal handles those
		operations.  Don't set it nonzero otherwise!

FS I&D LINE	if nonzero, tells TECO to try to use the insert and
		delete line operations to speed redisplay.
		Initialized nonzero if the terminal handles those
		operations.  Don't set it nonzero otherwise!

FS IBASE	the input radix for numbers not followed by "."
		(initially 8+2)

FS I.BASE	the input radix for numbers ended by ".".
		Initially 8.

FS IF ACCESS	(write-only) sets the access pointer of the
		current input file - the argument is the
		desired character address in the file.

FS IF CDATE	the creation date of the currently open input file.
		Arg and value are in numeric file date form -
		see FS FD CONVERT.

FS IF DEVICE	(read-only) is the device name of the current or
		most recent input file, as a numeric SIXBIT word.

FS IF DUMP	(ITS only) is the dumped-bit of the currently open input file.

FS IF FDB	(Twenex only) is used for reading and writing the file
		descriptor block of the current input file.
		<n>FS IF FDB returns word <n>, and <m>,<n>FS IF FDB
		sets it to <m>.

FS I FILE	(read-only) is the name of the current or most
		recent input file, as a string.  The format is like
		that of FS D FILE.

FS IF FN1	(read-only) is the first file name of the current or
		most recent input file, as a numeric SIXBIT word.

FS IF FN2	(read-only) is the second file name of the current or
		most recent input file, as a numeric SIXBIT word.

FS IF LENGTH	(read-only) the length, in characters, of the
		currently open input file;  or -1, if that length
		is unknown (because the file is on a device for
		which the fillen system call is unimplemented).
		Error if no file is open.

FS IF LINKP	(read-only) -1 if the currently open input file was
		reached by tracing a link; zero otherwise.  Always
		zero on Twenex.  On old versions of ITS it is also
		always zero.

FS IF MTAPE	(write-only) executes a .MTAPE call on the input file.
		The last argument is the .MTAPE operation.
		A preceding argument specifies the count (default is
		1).  It is not clear that this feature really works.

FS IF REAP	is the don't-reap bit of the currently open input file.

FS IF SNAME	(read-only) is the SNAME of the current or most
		recent input file, as a numeric SIXBIT word.

FS IF VERSION	(read-only) is the version number of the open input
		file, or -1 if there is none or its FN2 isn't numeric.	

FS IMAGE OUT	outputs arg as to the terminal as a character in
		superimage mode.  Returns no value.  If the argument
		is a string pointer, the contents of the string are
		output.

FS IN COUNT	is an old name for FS TYI COUNT.

FS INSLEN	length of last string inserted into the buffer with
		an "I", "G" or "\", or found with a search command
		or "FW".  FS INSLEN will be negative after a backward
		search or "FW" with negative arg.  See "FK" and "^F".

FS INVMOD	if nonzero, causes the mode line to be displayed in
		inverse video on terminals which support it.

FS JNAME	(read only) returns the JNAME of the job TECO
		is running in, as a numeric SIXBIT word, which
		can be converted into text by the F6 command.
		Note that the XJNAME is also available, and might
		be better for your purpose - see FS XJNAME.

FS JRN EXECUTE	(write only) opens a journal file for execution
		(replaying).  The default filenames are used.  With
		a colon modifier, the journal file being replayed is
		closed.
		See the section on journal files for more information.

FS JRN IN	(read only) is nonzero when a journal file is being
		replayed.

FS JRN INHIBIT if nonzero causes input to be taken from the terminal
		even though a journal file is being replayed.
		This is how FS JRN MACRO can offer the user a
		recursive ^R to fix things up after a quit.

FS JRN INTERVAL
		The value of this flag controls how often the journal
		file being written is updated on disk.  It is in the
		form of a number of commands.  It is initially 50.

FS JRN MACRO	This macro is called whenever TECO attempts to read a
		command from a journal file being replayed and finds a
		colon or a ^G character in the file.  The colon or ^G
		is passed to the macro as a numeric argument.  In the
		case of ^G, this macro should execute a ^R, and then
		quit by setting FS QUIT to -1.  In the case of a
		colon, the macro should read more characters from the
		file using FS JRN READ and act on them.  The format
		of these characters is not specified.

FS JRN OPEN	(write only) opens a journal file for writing.
		The default filenames are used.  With a colon
		modifier, the output journal file is closed.
		See the section on journal files for more information.

FS JRN OUT	(read only) is nonzero when a journal file is being
		written.

FS JRN READ	(read only) reads a character from the input journal
		file being replayed.  If there is none, it returns a
		random value.

FS JRN WRITE	(write only) outputs its argument, a character or
		string, to the journal file being written.  It does
		nothing if no journal file is being written.

FS LAST PAGE	(read only) set to -1 when an input file is opened;
		set to 0 as soon as the last char of the file
		is read in.  Saved by E[ - E].  Updated by
		FS IF ACCESS.  Thus, FS LAST PAGE; will exit an
		iteration if there is no more to be read.

FS LEDEFS	(initially 0) specifies the correspondence between
		definitions of ^R command characters and the
		function codes of the local editing protocol.
		Its value should be a qvector containing pairs
		of elements: alternating definitions and function
		codes.

		The definitions are macros or built-in functions,
		the same as would go in FS ^R CMAC.  The function
		codes are documented in the local editing protocol.

		FS LEDEFS can also have a string value.  In this
		case, when the value is wanted, the string is executed
		as a macro.  It should set FS LEDEFS to a qvector as
		described above.

		If FS LEDEFS is zero, the local editing protocol is
		not used.
		
FS LINES	(initially 0) determines the number of lines
		displayed by standard buffer display, and, for
		display terminals, the number of lines to use at all.
		0 => a full screen on displays, 2 lines on
		printing terminals.  <n> not zero => <n> lines.

FS LISPT	normally 0, this flag is set nonzero if TECO
		is started at 2 + its normal starting address.
		This is intended to indicate to TECO programs
		that passing of text between TECO and its
		superior is desired.

FS LISTEN	returns nonzero if there is input available to
		be read by "FI".  If given an arg, then if no
		input is available, the arg is typed out using
		FS ECHOOUT.

FS LOAD AV	(Twenex only) returns the system load average.

FS MACHINE	(read only) returns the name of the machine TECO is
		running on, as a numeric SIXBIT word, which can be
		turned into text with F6.  On ARPANET Twenices, this
		returns the local hosts hostname as a string.

FS MODE CHANGE	if nonzero means that FS MODE MACRO needs to be run
		eventually, to update the contents of ..J.  Whenever
		TECO is considering updating the mode line, depending
		on the value of this flag it may run FS MODE MACRO
		first.  That macro can recompute ..J to display the
		current state of things.  FS MODE CHANGE is set to
		zero before FS MODE MACRO is called.

		The precise condition is: if FS MODE CHANGE is
		positive, do run FS MODE MACRO.  If FS MODE CHANGE
		is negative, its absolute value is compared with twice
		the value of FS QP PTR;  if the latter is less, do
		run FS MODE MACRO.  Thus, if FS MODE MACRO sets FS
		MODE CHANGE based on the current value of FS QP PTR,
		it can arranged to be called again if control returns
		to an outer level on the stack.  See Q..J and FR.

FS MODE MACRO	if nonzero is the macro used to update Q..J when the
		mode line is about to be displayed.  See FS MODE
		CHANGE and Q..J and FR.

FS MODIFIED	is the current buffer's modified flag.  This flag is
		set nonzero whenever the buffer's contents are changed.
		It can be read or written by the user at any time.
		Each buffer has its own flag, but only the selected
		buffer's flag is accessible.  The intended use is for
		deciding whether a file must be written back to the
		disk.  When this flag is nonzero, a star appears at
		the end of the mode line.  See also FS READ ONLY and
		FS X MODIFIED and FS ^R STAR.

FS MP DISPLAY	(write only, ITS only) outputs text to the main program
		area in display mode (^P is special).  The argument may
		be a single character or a pointer to a string, whose
		entire contents are output.

FS MSNAME	is the user's working directory name (set up from
		the SNAME that TECO was given when it started up),
		as a numeric SIXBIT word, which can be converted into
		text by the F6 command.  On Tenex or Tops-20, this is
		the connected directory and writing into it is used
		to connect to another directory.

FS NOCEOL	if nonzero, signifies that the terminal in use does
		not have the clear-to-end-of-line function.
		Programs may wish to clear the screen at times if this
		flag is set, if they know that clearing the screen
		would result in faster redisplay afterward on such
		terminals.

FS NOOP ALTMODE
		if negative, altmode is always a noop as a command.
		If 0, altmode is an error as a command.
		If >0, always ends execution (as ^_ does).
		Initially -1.  The old treatment of altmodes
		was as if this flag were set to 1.

FS NOQUIT	gives the user control of TECO's ^G-quit
		mechanism.  See "^G".

FS OF ACCESS	(write-only) sets the access pointer in the
		output file.  The argument must be a multiple of
		5.  If the last output done did not end on a
		word boundary, rather than throwing away the
		remaining characters, an error occurs.
		FS OF LENGTH gives the length of the file, if you
		want to append to the end.

FS OF CDATE	the creation date of the currently open output
		file, in numeric file-date form.

FS O FILE	returns the filenames of the last output file
		explicitly closed, as a string.  The format is
		like that of FS D FILE.  The intended use is for
		finding out what version number was actually written.

FS OF LENGTH	returns the length in characters of the currently open
		output file, or -1 if it cannot be determined.

FS OF MTAPE	(write-only) does a .MTAPE call on the output file.
		The last argument is the .MTAPE operation.
		A preceding argument specifies the count (default is
		1).  It is not clear that this feature really works.

FS OF VERSION	(read-only) is the version number of the last output
		file closed, or -1 if that file's FN2 was not numeric.

FS OLD FLUSHED	Saves the value of FS FLUSHED when that is set to
		zero, on returning to ^R.  Thus, a ^R command can
		check FS OLD FLUSHED to see whether the previous
		command had output which was flushed.

FS OLD MODE	is the last ..J value actually displayed in the mode
		line.  Setting this flag to zero will force redisplay
		of the mode line.

FS OSPEED	The terminal's ouput line speed in baud, or 0 if the
		speed is not known.

FS OS TECO	returns the operating system TECO is running on,
		0 for ITS, 1 for 20X, 2 for 10X.

FS OUTPUT	if nonzero, suppresses output to the EW'd file.
		Output commands (P, etc) are errors.

FS PAD CHAR	(T(w)enex only) is the character to use for padding
		terminal output.  Normally it is a Rubout (177).  For
		some terminals a null (0) is better.  Some may set it
		to 0 automatically.  -1 means use an actual delay
		instead of padding characters.  This may make some
		terminals work better, and may be more efficient of
		CPU time; however, it will not work across networks of
		any sort.

FS PAGENUM	the number of formfeeds read (with non-atsign Y and A
		commands) from the input file since it was opened.

FS PJATY	this is negative if TECO has detected that the terminal
		was temporarily taken away from it.  A negative value
		here will cause the whole screen to be refreshed at the
		next opportunity.

FS PROMPT	the ASCII value of the prompt character
		(initially 38 for "&").  TECO will prompt
		on printing terminals only, whenever it is about
		to read from the terminal, and FS PROMPT is not 0.

FS PUSHPT	(write-only) pushes its argument on the "ring
		buffer of the pointer", but only if the
		argument differs from the value already at the
		top of the ring buffer.  See the ^V command.
		Note that TECO's top level loop does
		Q..IFS PUSHPT each time it is about to read in
		a command string.

FS QP HOME:
<n>FS QP HOME	returns a string which says where the q-register
		pdl slot <n> was pushed from.  <n> can be a
		nonnegative offset from the bottom of the stack or
		a negetive offset from the top.  The string will
		contain the name of the q-register of FS flag that
		was pushed, as, for example, "QA", "Q.^RX", "QFoo",
		"FS DFILE", or "*" (for a [(...) ).  "*" isn't really
		a q-register name, but what would be any better?

<n>:FS QP HOME	returns the address of the q-register which
		q-register pdl slot <n> was pushed from (for a variable,
		returns the name of the variable).  Such numbers
		will be equal for two slots pushed from the same place
		in the same TECO.  They are also used with F^G.

<n>@FS QP HOME
		turns a q-register address returned by :FS QP HOME
		into a string such as plain FS QP HOME would return.
		@FS QP HOME may be useful in for decoding the local
		q-register home numbers used with the F^G command.

FS QP PTR	the q-register pdl pointer (0 if nothing has been
		pushed, 1 if one q-reg has been pushed, etc.)
		<n> FS QP PTR sets the pointer to <n> if <n> > 0;
		adds <n> to it if <n> is negative.  It is illegal
		to increase the pointer value with this command.

FS QP SLOT	<n> FS QP SLOT  reads q-reg-pdl slot <n>.
		<m>,<n>FS QPSLOT  sets it to <m>.
		The first slot is numbered 0.
		If <n> is negative, it is treated as FS QPPTR+<n>.
		Thus, -1FS QP SLOT is the last slot pushed.

FS QP UNWIND	(write only!) like FS QP PTR but pops
		slots back into the q-reg's they were pushed from
		instead of simply decrementing the pdl poinetr.
		This unwinding is automatically done when an error
		is caught by an errset, and at the end of each
		command string, and by ^\ and FS ^REXIT.
		If Q..N is popped by this command, it is macroed
		first (see ..N).
		If <n> is negative, <n>FS QP UNWIND pops -<n> slots.

FS QUIT	^G-quit works by setting this flag negative.
		Whenever quitting is possible (see FS NOQUIT)
		and FS QUIT is negative, quitting will occur
		(and FS QUIT will be zeroed automatically).
		When TECO's quitting is inhibited, the user can
		test this flag explicitly to do his own special
		quitting.

<n>FS Q VECTOR	(a pseudoflag) returns a newly cons'ed up q-register
		vector, <n> characters long.  <n> should normally
		be a multiple of 5.  The contents are initialized
		to zeros, and the pointer is at the beginning
		of the qvector.  A q-vector is really a buffer, each
		word of which is marked by the garbage collector.
		See the section on data types.

FS RANDOM	reads or sets the ^Z command random number
		generator's seed.

FS READ ONLY	if nonzero makes it an error to modify the buffer.
		This flag is actually per-buffer, so when you set it,
		you set it only for the currently selected buffer.

FS REAL ADDRESS
		returns the value of BEG, the character address
		of the beginning of the current buffer.  Useful
		for communicating with other programs that need
		to be given addresses of data in their commands.
		Also useful for executing the buffer as PDP10
		code (do FS REALAD/5U0 M0;  the code should be
		position-independent, expect its address in
		accumulator 5, and start with SKIP (skip never)
		instruction.  Its argument will be in AC 3.).

FS REFRESH	if nonzero, is macroed each time TECO really
		clears the whole screen.  It is given no args
		and its values are ignored.  The screen will
		already have been cleared.
		When only part of the screen is in use (because FS
		LINES and FS TOP LINE call for that), the only time
		that TECO will clear the whole screen is when it is
		continued after having been stopped, the primary use
		of this hook is to put back on the screen whatever is
		supposed to be in the part of the screen not being
		actively used for display.
		Functions which remove part of the screen from
		active use can sometimes be entered recursively, each
		one removing a few more lines from use.  Each of them
		might supply an FS REFRESH to refresh those lines.
		What ought to happen is that all the FS REFRESHes are
		executed, in the order they were set up.  To make this
		happen, any function which sets up an FS REFRESH
		should save the previous contents elsewhere, and the
		new FS REFRESH should call the old one first thing.

FS REREAD	(usually -1) if nonnegative, FS REREAD is the
		9-bit TV code for a character to be re-read.
		Putting 65 into FS REREAD will cause the next
		"FI" command to return 65 (and set FS REREAD
		back to -1).

FS RGETTY	0 if printing console,
		otherwise equal to the tctyp word of the
		terminal.  However, it is better to decode the
		FS %TOCID, etc., flags than to decode FS RGETTY
		when trying to determine what kind of display
		the terminal is, and what functions it can perform.
		On Twenex it is actually allowed to set this variable
		in case the system is mistaken about your terminal
		type.  What it must be set to is the TECO internal
		type code for the terminal you are using;  and not all
		types are assembled in on any particular Twenex site.

FS RUB CRLF	if nonzero causes the initial definitions of
		^D, rubout and control-rubout to delete both
		characters of a CRLF at one blow, as if it were
		a single character.

FS RUB MACRO	if nonzero is a macro to be called to perform the
		^R commands Rubout and C-D when they have explicit
		numeric arguments.  The normal definitions of the
		two characters call this macro in that case.

FS RUNTIME	(read-only) TECO's runtime in milliseconds.

FS SAIL	if nonzero, the terminal is assumed to be able
		to print non-formatting control chars as 1-space
		graphics.  TECO outputs them as they
		are instead of outputting an ^ and a non-ctl char.
		Terminal initialization zeros this flag if the
		terminal's TOSA1 bit is 0
		(this bit is set by :TCTYP SAIL).

FS S ERROR	if 0, as it is initally, a failing serach within an
		iteration or a ^P-sort key is not an error - it
		simply fails to move the pointer.  If not 0, such
		searches cause sfl errors like all other searches.

FS SHOW MODE	if nonzero, causes FR to type out the mode line
		(Q..J) on printing terminals, if it has changed
		since last printed.  Has no effect on displays.

FS S HPOS	(read only) is the horizontal position of point,
		taking control characters, tabs, CRs, etc. to appear
		the way they are currently being displayed, but
		assuming an infinitely wide line.

FS S STRING	is the default search string (which S will use),
		as a string pointer.  Do G(F S STRING) to insert
		it in the buffer.  This flag is most useful for
		pushing and popping.

FS STEP MACRO	(normally 0) if nonzero and numeric, TECO displays
		the buffer and waits at the beginning of every line
		of the program.  See ^M for details.
		If it is a string, it is executed at the beginning
		of every line in the program.
		Exception:  if a macro begins with a "W", stepping
		is inhibited at the beginning of that macro.
		The macro can inhibit stepping throughout itself
		by starting with "W 0F[STEP MAC".
		Exception:  stepping is also done whenever a
		successful conditional begins.

FS STEP DEPTH	restricts the macro call stack levels at which
		stepping is done.  If FS STEP DEPTH is -1, as it is
		initially, stepping goes on at all stack levels.
		Otherwise, the value of FS STEP DEPTH is the largest
		macro call depth (FS BACK DEPTH value) for which
		stepping can take place.

FS SUPERIOR	if nonzero is the macro to be called when TECO's
		superior makes a request to insert text into TECO.
		See "How Superiors Can Put Text into TECO" below.

FS S VALUE	the value stored by the last search command
		(0 if search failed;  else negative, and minus
		the number of the search alternative which was
		actually found).

FS TAB WIDTH   (initially 8) is the spacing between tab stops
		for display of tab characters in the buffer.
		Although the system always uses 8-character tab
		stops, TECO handles output of tabs itself and can
		output with any interval between tab stops.

FS TOP LINE	is, on display terminals, the number of the first
		line on the screen that TECO should use.
		Normally 0, so TECO starts output at the top of
		the screen.

FS TRACE	nonzero iff TECO is in trace mode.  See "?".
		1F[ TRACE is a good way to start tracing temporarily.

FS TRUNCATE	says what to do with long lines of type-out.
		Negative => truncate them.
		Positive or zero => continue them to the next line.
		Entering ^R sets this flag to 0.

FS TTMODE	(initially 0) non-zero tells TECO that normal
		buffer display should display on printing terminals.
		(if there is a user buffer display macro, this
		flag has no effect unless the macro checks it)

FS TTY FCI	(Twenex only) if non-zero, we assume that the terminal
		has an Edit key, and that the 200 bit in incoming
		characters should be interpreted as "Meta".  If this
		flag is zero, then the 200 bit is ignored on the
		assumption that it is a parity bit.

FS TTY INIT	(no argument or value) causes TECO to reexamine
		the system's terminal description and reset various
		flags, and the cursor in ..A, appropriately.
		Then, FS TTY MACRO is executed if nonzero.  It can be
		used to digest the tty characteristics and set flags
		accordingly.  FS TTY INIT is run automatically when
		TECO is started or restarted.
		In detail, ..A is set to "-!-" on printing terminals,
		and to "/\" on displays (but "^A^B" on imlacs).
		FS RGETTY is set up to be 0 on a printing terminal,
		nonzero otherwise.  FS VERBOSE is set equal to
		FS RGETTY.  FS TTYOPT is read in from the system.
		FS ^H PRINT is zeroed unless the terminal
		can backspace and overprint;  FS ^M PRINT is zeroed
		unless the terminal can overprint.
		FS SAIL is set nonzero if the %TOSA1 bit is set in
		TTYOPT (this is the bit ":TCTYP SAIL" sets).
		FS LID is set nonzero if the %TOLID bit is set in
		TTYOPT.  %TOLID says that the terminal can insert and
		delete lines.
		FS WIDTH and FS HEIGHT are read in from the system.
		FS ECHOLINES is set to 0 on printing terminals;
		1/6 of the screen size on displays.
		FS D WAIT is set to -1 if FS OSPEED is positive and
		600 or less;  to zero, otherwise.

FS TTY MACRO	is a macro called by FS TTY INIT to give the user a
		chance to alter TECO flag settings based on the
		newly-gobbled terminal characteristics.
		If you don't like the way TECO initializes certain FS
		flags (namely FS ECHOLINES, FS TRUNCATE, FS
		VERBOSE, FS WIDTH, FS ^HPRINT, FS ^MPRINT, and FS
		SAIL) in FS TTY INIT, this macro can change them.
		Pure strings should NOT be used for this; when a saved
		EMACS environment is restored, FS TTY MACRO$ is
		processed before even the standard libraries are
		loaded in.  So you must create the string with :I, X
		or :G.  You can make an impure copy of a pure string
		in Q0 with :I00.

FS TTYOPT	(read-only) the TTYOPT word for the terminal.
		However, if you think you want to use this flag, see
		FS %TOCID, etc., first.

FS TTY PAGE MODE
		(Twenex only) if negative, prevents TECO from leaving
		page mode, so that ^Q and ^S are still used for flow-control.
		Note that this means they are not available as commands.

FS TTYSMT	(read-only) the TTYSMT word for the terminal.

FS TYI BEG	is the value which FS TYI COUNT had on the last time
		through the main ^R command loop.  In other words, it
		is the number of terminal input characters so far, not
		including the last or current ^R command.  Commands to
		set the numeric argument do not count as real commands
		in this regard;  they are grouped with the commands
		for which they set the argument.
		When ^R is reading the first character of a command,
		FS TYI COUNT has been incremented already, so it
		equals FS TYI BEG+1.

FS TYI COUNT	is the number of characters read from the terminal so
		far.

FS TYI SINK	if nonzero holds a macro to be called every time a
		character is read from the terminal (not from FS
		REREAD or from FS TYI SOURCE).  The character read,
		in the 9-bit code, is passed as a numeric argument to
		the macro.  The macro should return a single value,
		for the sake of ^R display updating.  FS TYI COUNT
		and FS TYI BEG may be of use in the macro.

FS TYI SOURCE	if nonzero holds a macro to be called to obtain
		"terminal input".  This is instead of reading from the
		keyboard.  The macro should put the input character
		it wishes to provide into FS REREAD and then return.
		The macro should return a single value, for the sake
		of ^R display updating, but this value is not used in
		any other way.  That FS TYI SOURCE is nonzero is
		taken to imply that input is available, so display
		updating is suppressed.  To cause displaying to be
		done, you must bind FS TYI SOURCE to zero and then
		request it.  FS TYI SOURCE is set to zero by anything
		that discards keyboard input (errors, quits, etc).

FS TYO HASH	<vpos>FS TYO HASH returns the hash code of the screen
		line at position <vpos>.  <n>,<vpos>FS TYO HASH sets
		the hash code of that screen line to <n>.  This is
		useful primarily for -1,<vpos>FS TYO HASH which will
		force that screen line to be redisplayed.

FS TYO HPOS	(read-only) while typeout is in progress (FS TYPEOUT
		nonnegative), holds the current typeout horizontal
		position, in which the next typed character will appear.

FS TYO VPOS	(read only) while typeout is in progress (FS TYPEOUT
		nonnegative), holds the current typeout vertical
		position, in which the next typed character will appear.

FS TYPEOUT	is -1 if typeout has not been happening recently,
		so typeout starting now would appear at the top of
		the window.  FS TYPEOUT is not -1 when typeout was
		the last thing to happen and any more typeout will
		appear after the previous typeout.  :FT types at
		the top of the window by putting -1 in FS TYPEOUT
		before typing.

FS U HSNAME	is used to determine a user's hsname.
		<user> FS U HSNAME (where <user> is sixbit) returns
		<user>'s hsname in sixbit.  <its>,<user> FS U HSNAME
		returns his hsname on machine <its> (sixbit).

FS UINDEX	(read only) The user index of the TECO job.

FS U MAIL FILE	is used to determine the filename (including the
		machine) of a user's mail file.  <user>FS U MAIL sets
		the default filename to the name of <user>'s mail file
		(where <user> is in sixbit).  <its>,<user>FS U MAIL
		sets them to the name of his mail file on machine
		<its> instead of his normal mail-receiving machine.

FS UNAME	(read only) returns the UNAME of the job TECO is
		running in, as a numeric SIXBIT word, which can be
		converted into text by the F6 command.  See also
		FS XUNAME and FS MSNAME, one of which might be
		better for your purpose.

FS UPTIME	(read only) returns time system has been up, in 30'ths,
		(milliseconds on Twenex).

FS UREAD	(read-only) -1 iff an input file is open, else 0.
		Once an input file is opened, it remains open until
		"EC", "@Y", "@A", "EE" or "EX" is done.

FS UWRITE	(read-only) -1 if an output file is open, else 0.

FS VAR MACRO	if nonzero enables the feature whereby setting a named
		variable can run an arbitrary macro to make changes in
		other data structures.  These data structures will
		appear automatically to reflect the value of the
		variable.

		To use this feature, allocate three words per variable
		in the symbol table in Q..Q.  The macro to call is the
		contents of the third word, but only if it is a string
		which starts with a "!".  If that is the case, it will
		be called with the new value of the variable as its
		argument.

FS V B		is the distance between the real beginning of the
		buffer and the virtual beginning.  See FS BOUNDARIES,
		but unlike that flag, FS V B can be pushed and popped.
		B has the same value as FS V B.

FS VERBOSE	if not 0, TECO will print the long error message
		of its own accord when an error occurs.
		Otherwise it will print only the 3-char code
		and the long message must be requested by typing
		^X.  Initially 0 except on displays.

FS VERSION	(read-only) the current TECO version number

FS V Z		is the distance between the virtual end of the buffer
		and the real end - the number of characters past the
		virtual end.  See FS BOUNDARIES for more info, but
		note that FS V Z can be pushed and popped.
		Z's value is FS Z-(FS V Z).

FS WIDTH	width of terminal's screen or paper, in characters.

FS WINDOW	the number of the first character in the
		current display window, relative to the virtual
		beginning of the buffer (that is, FS WINDOW+B
		is the number of that charcter).
		 FS WINDOW+BJ (FS HEIGHT-(FS ECHOLINES)/2)L
		will put the pointer in the middle of the window
		(usually).  Setting FS WINDOW will make TECO try
		to use the window specified.  However, if the
		constraints of FS %TOP and FS %BOTTOM are not
		met, TECO will choose another window rather than
		use the specified one.

FS WORD	gets or sets words in the current buffer.  This
		flag makes it possible for TECO programs to
		edit binary data bases.
		<n>FS WORD returns the contents of the word
		containing character <n>;  <val>,<n>FS WORD
		sets that same word.
		When handling binary data, it is unwise to
		insert or delete characters other than in units
		of five, on word boundaries.  The way to delete
		a word is to delete its five characters;  insert
		a word with 5,0I (this does not set the low bit!
		You cannot count on the word to contain 0 unless
		you store into it with FS WORD).

		To read in a file of binary data, FY should be 
		used, since Y might pay special attention to the
		characters in the file.  Copying out of another
		buffer with G works, provided the transfer starts
		and ends on word boundaries (blt is used).
		For writing out binary data, use @HP.  The @ says
		not to clear the low bits.  Use HP rather than PW
		PW since PW may add a ^L.  EF is OK for closing the
		file - it will add no padding if it is done at a
		word boundary.

FS XJNAME	(read only) returns the XJNAME of the job TECO is
		running in, as a numeric SIXBIT word, which can be
		converted into text by the F6 command. The XJNAME is
		"what the JNAME was supposed to be", so if you want
		your init file to do different things according
		to how TECO was invoked, you should use the XJNAME
		rather than the JNAME.  On Tops-20, this will be the
		first atom of the command line, if there is one; this
		is useful for the same reason.

FS X MODIFIED	is completely analogous to FS MODIFIED in how TECO
		treats it.  However, the user can clear either one
		without changing the other.  They can be used to
		detect the occurrence of modifications since two
		different types of events (such as, real saves and
		auto saves).  FS X MODIFIED does not affect the star
		in the mode line; only FS MODIFIED does.
	  
FS X PROMPT	On printing terminals, all commands that type out
		first print and zero FS X PROMPT if nonzero (using
		FS ECHO OUT).  Do not try to use it on displays.

FS XUNAME	(read only) returns the XUNAME of the job TECO
		is running in, as a numeric SIXBIT word, which
		can be converted into text by the F6 command.
		The XUNAME is "who the user really is".
		For example, it is what TECO and other programs
		use to decide whose init file to use.

FS Y DISABLE	controls treatment of Y command.
		0 => Y is legal.
		1 => Y is illegal (gives "DCD" error).
		-1 => Y is always treated as @Y.

FS Z		(read only) the number of characters in the buffer.
		Will differ from value of Z command when virtual
		buffer boundaries don't include the whole buffer.
		This is Z-BEG (see "buffer block").

	note: in the names of the following flags,
	"^" represents caret, not a control character.
	Control characters cannot be part of FS flag names.

FS ^H PRINT	controls how ^H is typed out.
		Negative => actually backspace (and overprint),
		otherwise type the ^H as caret-H.
		FS TTYINIT and G cause this flag to be zeroed
		if the terminal cannot handle overprinting.

FS ^I DISABLE	controls the treatment of the tab character
		as a TECO command.
		0 => ^I is an insert command (see Tab for details).
		1 => ^I is illegal (gives "DCD" error).
		-1 => ^I is a no-op.

FS ^L INSERT	(initially 0) if 0, formfeeds in files that terminate
		Y commands' reading, are thrown away, and the P and PW
		commands output a formfeed after the buffer.
		If FS ^LINSERT is nonzero, formfeeds read from files
		always go in the buffer, and P and PW never output
		anything except what is in the buffer.
		Either way, a Y and a P will write out what it
		reads in.

FS ^M PRINT	says when a stray CR or LF should be typed out
		as one, as opposed to being printed as "^M" or "^J".
		Possible values and initialization like FS ^H PRINT

FS ^P CASE	if nonzero, ^P sort ignores case (lowercase
		letters sort like the corresponding uppercase).

FS ^R ARG	is the explicit numeric argument for the next ^R
		mode command, or 0 (not 1!) if there was none.

FS ^R ARGP	contains two bits describing this ^R-command's
		argument:
		  bit 1.1 (1) is set if any argument was specified
			(either numerically or with ^U).
		  bit 1.2 (2) is set if a numeric argument was
			specified.  If this bit is 0, the contents of
			FS ^R ARG are ignored, and 1 is used instead.
		  bit 1.3 (4) is set if the argument, as specified by
			the other bits, should finally be negated.

FS ^R CCOL	the comment column, for ^R's comment mode.

FS ^R CMACRO	<n>FS ^RCMAC gets the ^R-mode definition of
		the character whose ASCII code is <n>.
		<m>,<n>FS ^RCMAC sets it to <m>.
		<n> should be an ASCII code;  it will be converted
		to 9-bit TV code which is what is actually used
		to index the ^R-mode dispatch table.
		If you wish to supply a 9-bit code yourself, use
		"@FS ^RCMAC" which skips the conversion.
		Also, these definitions may now be referred to
		as q-regs in all the q-reg commands - see "Q".
		The definition is either a built-in command or
		a user macro.  In the former case, the
		definition is a positive number of internal
		significance only.  However, built-in
		definitions may be copied from one character
		to another using FS ^RCMAC.  For a character to
		be a user macro, its definition must be one of
		the funny negative numbers which are really
		string pointers.  They can be obtained from
		strings by applying the "Q" command to a q-reg
		that contains text.  For example, to make the
		definition of the character " " be the string
		which is at the moment in q-reg A, do
		"QA,^^ FS ^RCMACW".  "QAU^R " is equivalent.
		To copy the definition of
		"A" into the definition of rubout, do
		"^^AFS ^RCMAC,127FS ^RCMACW".  This will make
		rubout self inserting (unless "A" had been
		redefined previously).

FS ^R DISPLAY	if nonzero is macroed every time ^R is about to do
		nontrivial redisplay (anything except just moving
		the cursor).
		If the evaluation alters the needed redisplay
		(either by returning 0 or 2 values to ^R, or by
		doing some of the redisplay with ^V) then ^R will
		take note. If a FS ^R DISPLAY returns no values,
		it will force a full redisplay, thus effectively
		disabling ^R's short-cuts, so beware.
		If FS REFRESH is nonzero, then it will be called
		in addition to FS ^R DISPLAY, at those times when
		the whole screen is being cleared.

FS ^R EXIT	(write-only) exits from the innermost ^R invocation.
		Pops q-regs pushed within that ^R level, and ends
		iterations started within it.

FS ^R ECHO	1 => characters read in by ^R should not be echoed.
		0 (the default) => they should be echoed only on
		printing terminals.
		-1 => they should be echoed on all terminals.
		Note that this "echoing" is explicit typeout by TECO.
		System echoing is always off in ^R mode, currently.
		Also, rubout is not echoed on printing terminals.
		However, FS ^R RUBOUT on a printing terminal when
		FS ^R ECHO is <= 0, types out the char being
		deleted.

FS ^R EC SD	is used to allow system echoing to echo the Space
		character when some sort of auto-fill feature is in
		use and Space does not have the definition of a normal
		self-inserting character.  This flag should be set to
		the definition which Space has when it does
		auto-filling.  When Space has this definition, it will
		be considered safe to let the system echo Spaces just
		as if they were ordinary self-inserting characters;
		however, to prevent auto-fill from being lost, system
		echoing will be allowed only as far as the column
		which is the value of FS ADLINE, which should be set
		no greater than the column at which auto-fill starts
		to do something nontrivial.  The definition of Space
		will not actually be executed when the space is echoed
		by the system, but for an auto-fill which isn't going
		to break the line this should make no difference.

FS ^R ENTER	is macroed (if nonzero) whenever ^R is entered
		at any level of recursion.

FS ^R EXPT	is the ^U-count for the next ^R-mode command.

FS ^R H MIN	(read only) is the hpos of the first change on the
		screen needing redisplay.  It will be a large positive
		number if no redisplay is required.  If FS WINDOW is
		negative then the value of this flag doesn't matter.

FS ^R HPOS	the current horizontal position of the cursor
		in ^R mode.  Not updated when the pointer moves,
		unless ^R gets control back or @V is done.

FS ^R INDIRECT	given a 9-bit character, traces ^R alias-definitions
		to find the character it is equivalent to.  If given
		a character that isn't an alias, returns the same one.
		Thus, 311. FS ^R IND returns 11.

FS ^R INHIBIT	when nonzero prevents ^R from updating the display.
		If FS ^R INHIBIT is zeroed again, all the pending
		updating will be done at the next opportunity.

FS ^R INIT	<ch>FS ^R INIT returns the initial definition of
		the character whose ASCII code is <ch>;  in other
		words, <ch>FS ^R INIT always returns what
		<ch>FS ^R CMACRO initially returns.
		The atsign modifier works for FS ^R INIT just
		as it does for FS ^R CMACRO;  it says that the
		arg is a 9-bit code rather than ASCII.

FS ^R INSERT	the internal ^R-mode insert routine's user interface.
		It takes one argument - the ASCII code for the
		character to be inserted.  This command itself take
		care of notifying ^R of the change that is made, so
		when returning to ^R this change should not be
		mentioned in the returned values (so if this is the
		only change made, return 1 value).
		This command is very sensitive;  if the buffer or
		even "." has changed since the last time ^R was
		in progress or an @V was done, it may not work.
		Its intended use is in macros which, after thinking,
		decide that they wish only to insert 1 or 2
		characters (such as a space-macro which might
		continue the line but usually inserts a space).

FS ^R LAST	holds the most recent character read by any ^R
		invocation (quite likely the one being processed
		right now).  Commands that wish to set up arguments
		for following commands should zero FS ^R LAST,
		which tells ^R not to flush the argument when
		the command is finished.

FS ^R LEAVE	is macroed (if not zero) whenever ^R returns
		normally (including FS ^R EXIT but not throws
		that go out past the ^R).

FS ^R MARK	holds the mark set by ^T in ^R mode, or -1 if
		there is no mark.

FS ^R MAX	holds the maximum number of characters of insertion
		or deletion that will be printed out by ^R on
		a printing terminal.  Larger changes will cause ^R
		to echo the command instead of displaying its effect.
		The default value is 50 .

FS ^R MCNT	the counter used by ^R to decide when to call
		the secretary macro.  It starts at FS ^R MDLY and
		counts down.

FS ^R MDLY	sets the number of characters that should be read
		by ^R mode before it invokes the secretary macro
		kept in q-register ..F .  Characters read by
		user macros called from ^R are not counted.

FS ^R MODE	(read-only) non-zero while in ^R-mode.

FS ^R MORE	if positive, --MORE-- is used for ^R-mode display
		instead of the usual --TOP--, --BOT-- and --nn%--.
		This is useful in command environments where Space
		means "show the next screenfull".
		If negative, then neither --MORE-- nor --TOP-- is
		displayed.  This may be desirable on slow terminals,
		and also allows ..J to make use of the full width of
		the mode line.

FS ^R NORMAL	all "self-inserting" characters in ^R mode are
		really initially defined to go indirect through
		this word, if it is nonzero.  If it is zero, as
		it is initially, the default definition is used
		for such characters.

FS ^R PAREN	holds a function to be executed whenever a
		self-inserting character is inserted whose Lisp syntax
		is ")".  The idea is that this function will point out
		the location of the matching open parenthesis.  If FS
		^R PAREN is zero, the feature is turned off.

FS ^R PREVIOUS	holds the previous (second most recent) command
		read by ^R's command loop, not counting argument
		setting commands.

FS ^R REPLACE	if nonzero puts ^R in "replace mode", in which
		normal characters replace a character instead of
		simply inserting themselves.  Thus, the character
		A would do DIA instead of just IA.  There are
		exceptions, though;  a ^H, ^J, ^L or ^M will not
		be deleted, and a tab will be deleted only if
		it is taking up just one space.  Also, characters
		with the meta bit set will still insert.
		Replace mode actually affects only the default
		definition of "normal" characters.  Characters which
		have been redefined are not affected, and if
		FS ^R NORMAL is nonzero no characters are affected
		(unless the user's definitions check this flag).
		Making FS ^R REPLACE positive has the additional
		effect of forcing all meta-non-control characters
		to be come normal, suppressing their definitions.

FS ^R RUBOUT	the internal ^R rubout routine's user interface.
		Takes 1 arg - the number of characters to rub out.
		This command is very sensitive and may fail to work
		if the buffer or "." has been changed since the
		last time ^R was in control, or an @V, FM,
		FS ^R RUB or FS ^R INSERT was done.  Its intended
		use is for macros which, after thinking, decide
		to do nothing but rub out one character and
		return;  it gives extra efficiency but only when
		rubbing out at the end of the line.

FS ^R SCAN	if nonzero causes ^R commands,
		when using a printing terminal, to try to imitate
		a printing terminal line editor by printing the
		characters they insert/delete/move over.
		FS ^R ECHO should be 1, to avoid double-echo.
		See also FS ^R TTM1 for how to customize this.

FS ^R STAR	if nonzero, a star appears in the mode line if the
		buffer is modified.  It is nonzero by default.

FS ^R SUPPRESS	(initially -1) nonnegative => builtin ^R-mode
		commands are suppressed, except for rubout,
		and user-defined commands are suppressed unless
		their definitions begin with "W".  Suppressed
		command characters become self-inserting.  The char
		whose 9-bit value is in FS ^R SUPPRESS is the
		unquoting char.  It reenables suppressed commands
		temporarily by setting FS ^R UNSUPPR to -1.  If
		FS ^RSUPPRESS is > 511, there is no unquote char.

FS ^R THROW	returns control to the innermost invocation of
		^R.  This is different from FS ^R EXIT, which
		returns control FROM that invocation.

FS ^R TTM1	if nonzero, will be called to handle display on
		printing terminals of operations TECO itself does not
		know how to display.  This includes operations that
		return no values to ^R, operations which only move the
		cursor but move it a long way, and operations which
		modify a region of the buffer which is longer than FS
		^R MAX or does not end at point.

		When an operation returns no values to ^R, FS ^R TTM1
		is called with no arguments.

		When an operation moves the cursor a long distance, FS
		^R TTM1 is passed one argument, which is the old
		cursor position.

		When an operation modifies a region which ^R itself
		does not know how to display for, FS ^R TTM1 receives
		two arguments, which delimit the region changed.
		These two argumemnts will not necessarily be in order.

		If FS ^R TTM1 does not output anything, then the
		command is echoed by ^R when FS ^R TTM1 returns.

		If FS ^R SCAN is zero, redisplay is not attempted on
		printing terminals and this macro is not called.

FS ^R UNSUPP	(initially 0) actually, builtin commands are
		suppressed only if this flag and FS ^RSUPRESS
		are nonnegative.  However, this flag is zeroed
		after each command except ^U and ^V.  Thus,
		setting this flag to -1 allows one builtin comand.

FS ^R V MIN	(read only) is the vpos of the first change on the
		screen needing redisplay.  It will be a large positive
		number if no redisplay is required.  If FS WINDOW is
		negative then the value of this flag doesn't matter.

FS ^R VPOS	the ^R-mode cursor's vertical position.

FS _ DISABLE	controls treatment of the "_" command.
		If 0 (the default), "_" is "search-and-yank"
		 as it originally was.
		If 1, "_" is illegal (gives "disabled command" error).
		If -1, "_" is treated like "-"
		 (good on memowrecks).

FT	types its string argument.
:FT	similar, but always goes to top of screen first (actually,
	to the line specified by FS TOP LINE).
@FT	similar to FT, but types its argument in the echo area
	rather than the display area.  Characters are typed normally,
	in ITS ASCII mode, rather than as they would echo, so to do
	a CRLF you need a CR and a LF.
@:FT	like @FT, but types the argument only if no input is
	available (FS LISTEN would return 0).  If input is
	available, the argument is skipped over and ignored.

FU	a list manipulating command whose main use
	is in <arg>FUL, which moves up <arg> levels of parentheses.
	<arg>FU where <arg> is positive returns a pair of args for
	the next command, specifying the range of the buffer from .
	Moving rightward to the first place <arg> levels up.
	If <arg> is negative, it moves left -<arg> levels up.

FV	displays its string argument.
:FV	types its string argument, then clears whatever is
	left of the screen.

FW	similar to FL but hacks words instead.
	A word is defined as a sequence of non-delimiters.
	Initially, the non-delimiters are just the squoze
	characters but the user can change that - see q-reg ..D.
	This command returns a pair of args for the next one.
	Also, FW sets FS INSLEN equal to the length of the
	last word moved over.
	Main uses: FWL moves right one word,
	-FWL moves left one, FWK deletes one word to the right,
	FWFXA deletes and puts in q-reg A,
	FWFC converts one word to lower case.

:FW	similar to FW but stops before crossing the word instead
	of after.  Thus, :1FWL moves up to before the next
	non-delimiter.  :2FWL is the same as 2FWL-FWL.
	:FW sets FS INSLEN to the length of the last inter-word
	gap crossed.

@FW	like FW, but finds LISP atoms rather than words.
	Understands slashes and vertical bars but not comments.

FX	like X and K combined.  "3FXA" = "3XA 3K".
	The atsign flag causes appending to the q-reg, as for X.

FY	insert all that remains of the current open input file
	before point.  Error if no file is open.  The input data are
	unaltered;  no attempt is made to remove padding or ^L's.
	If the transfer is on a word boundary in the file and in
	the buffer, word operations will be used, so this command
	is suitable for use with binary data.
	The input file is not closed - use EC for that.

<n>FY	like FY, but inserts at most <n> characters, or until EOF,
	whichever comes first.  Note that <n> characters of
	space are always needed, even if the file is not really that
	long;  thus, 1000000FY to read in the whole file will
	not work.  The input file is not closed.

FZ	(T(w)enex only) manipulates inferior forks.  This command does
	not work on Tenex but could probably be made to work.
FZ	resumes the inferior exec fork, creating it if there is none.
	There is at most one "exec fork" since FZ will not create one
	if one already exists.
0FZ<string>
	resumes the inferior exec fork, creating it if there is none.
	The "string" is placed in the Rescan buffer for use by the exec.
-1,0FZ<string>$
	creates and starts a new exec fork placing the "string" in
	the Rescan buffer.  This new exec fork replaces any existing
	one to become "the" exec fork, which following FZ's with no
	argument will resume.  There are no defaults for the filename.

FZ<filespec> <string>
	creates and starts a non-exec fork by loading the specified
	file (no defaults!).  The "string" is placed in the Rescan
	buffer in the same manner EXEC uses.  Ie. "FILENAME string".
	The FS FORK JCL flag can also be used to set the command
	line, it is just the "string" part.
	After the fork returns, the FZ command returns a positive
	number which is the TECO index of the fork.
	This index can be used to resume or kill the fork:
<i>FZ<string>
	resumes the fork of index <i>, and waits for it to return.
	The string is placed in the Rescan buffer.  A pre-comma
	argument may be given:
	-1 -- signifies that the Rescan buffer is to be preset
	      to read.
	0  -- start the fork at its primary entry location.
	n  -- start the fork at location n in the entry vector.
-<i>FZ
	kills the fork of index <i>.  Indices are always positive.
@FZ...	is like FZ but does not clear the screen on return.

F[<flag>
	pushes the value of FS<flag> on the q-reg pdl,
	so that it will be restored on unwinding.
<arg>F[<flag>
	pushes the flag and sets it to <arg>
<ch>F[ ^R CMACRO
	pushes the definition of character number <ch>.
<arg>,<ch>F[ ^R CMACRO
	pushes the definition of character number <ch> and sets it.

F_	this command has the same meaning that _ normally has;
	namely, search for a string arg and keep yanking till
	end of file.  However, this command works regardless of
	the setting of FS _DISABLE

F]<flag>
	pops from the q-reg pdl into FS <flag>.

<ch>F] ^R CMACRO
	pops from the q-reg pdl into the definition of character
	number <ch>, and returns the old definition.

F~	compares strings, ignoring case difference.  It is just
	like F= except that both strings are converted to upper
	case as they are compared.

G<q>	insert in buffer to left of pointer the text in q-reg <q>.
	If q-reg specified contains # rather than text, decimal
	representation thereof will be inserted.  If the q-reg
	contains a buffer the gap in the buffer may have to be
	moved before the G can be done.
	FS INSLEN is set to the length of the inserted text.
<m>,<n>G<q>
	insert only a part of the text in the q-reg;  specifically,
	the range from <m> to <n>-1 inclusive.  This feature works
	only for q-regs containing text;  if a q-reg holds a number
	the whole q-reg will be inserted despite the args.
:G<q>	returns a copy of the string in q-register <q>.
<n>:G<q>
	returns as a number the character at position <n> in the
	string in <q>.  It is an error if <n> is negative or >= the
	length of the string.
<m>,<n>:G<q>
	makes a substring of the string in <q>, taking the
	characters starting form position <m> and stopping before
	character <n>.  This substring is returned as a value.
	FS INSLEN is set to the length of the substring.

H	equivalent to B,Z;  i.e., specifies whole buffer
	(or all within the virtual boundaries if they're in use)
	to commands taking two args for character positions such as
	K, T, or V.

I	if no arg, insert following string arg (terminated by Altmode)
	into the buffer before point.  The characters are inserted at
	the pointer, and the pointer is left after the characters.
	If preceded by an atsign ("@"), then the
	following char is the delimiter, and the next un-quoted
	and un-delimiter-protected occurrence of that character
	is what ends the string argument.
	e.g., @I/text/.
	The length of the inserted string is kept in FS INSLEN
	(see the "FK" and "^F" commands).
:I<q>	takes a q-register name immediately after
	the I and inserts into that q-reg,
	replacing the previous contents.
	Atsign works as with the
	normal I command, with the delimiter
	following the q-reg name.
	FS INSLEN is not set by :I.
	Self-inserting chars will not take
	the colon modifier.
<n>I	inserts the character with ASCII code <n>.
<n>:I<q>
	puts the character in a string in q-reg <q>.
<m>,<n>I
	inserts <m> copies of the character with ASCII code <n>.
<m>,<n>:I<q>
	puts <m> copies of the character in a string in q-reg <q>.
J	position pointer after argth char
	in buffer.  If no arg, arg=B (usually 0).

:J	is to J as :S is to S.

K	if no arg or one arg, kill chars
	from pointer to argth line feed
	following.  (no arg, arg=1;  negative
	arg, back up over 1-arg line feeds,
	space over last line feed found, and
	kill from there to pointer.  A colon
	after the arg will move back over
	carriage return+linefeed before deleting.
	If no carriage return exists,
	TECO will only move back one character.
	There is an implicit line-feed at the end of 
	the buffer.  (note: this is the action of colon
	for all commands which take
	one or two args like K.)
<m>,<n>K
	kills characters <m> through <n>-1.
	The pointer is moved to <m>.

<n>@K	is like <n>K, but only LF's preceded by CR's are
	recognized.

L:
<arg>L	move to beginning of <arg>th line after
	pointer (0L is beginning of current
	line.).  Colon acts as in the K command.
	Note that :L moves to end of current line
	0:L moves to end of previous line
	and -:L moves to end of line before 
	previous line
<m>,<n>L is the same as <m>+<n>-.J

<arg>@L
	is like <arg> L but only CRLFs are recognized, not stray LF's.

M<q>	calls the function in q-reg <q>.  If <q> contains
	a string or buffer, its contents are "macroed" -
	that is, treated as TECO commands.
	If <q> contains a number, that number should be the
	initial definition of some ^R-mode character;
	that "built-in" function will be called.
	Built-in finctions take 1 arg;  user macros,
	0, 1 or 2, which they may access using "^X" and "^Y"
	(or, more winningly, with the F^X and F^Y
	commands).  The macro may read string arguments using
	the ^]^X construction;  such arguments should be
	supplied after the M command.
	Note that if you macro a buffer, you may screw yourself
	if while that buffer is executing you either
	modify its contents or throw away all pointers to it.
	If the q-register specified in the M command is a
	^R-mode character definition (as in M.^RX), the code
	for that character is put in Q..0 in case the definition
	looks at it there.

@M	is a variant of M that fools the called macro into thinking
	that it was called directly from ^R mode.  One effect is that
	the default postcomma numeric argument is 1, not 0.  To be
	precise, if a macro is called with @M and no arguments, then
	inside that macro F^X will still say that there are no
	arguments, but if ^Y is used anyway its value will be 0, not 1.

	Another effect is that if the called macro uses the F^K
	command to read a string argument, the argument will be read
	from the terminal.  If a macro called with @M does a :M, the @
	flag will be passed along to the macro called by the :M;
	however, this will not affect the argument passed by the :M,
	which will be determined solely by the code in the caller
	(including whether the caller explicitly says @).

:M	is a tail-recursive form of macro call; it is similar to a M
	followed by a ^\, for some purposes.  However, with :M the
	caller is really no longer present on the macro call stack.
	If the called macro reads a string argument, it will come not
	out of the caller, but out of the caller's caller.

N	does search (see S) but if end of
	buffer is reached does P and
	continues search.

O<tag>	sends command execution to char after
	the occurrence of <tag> as a label ("!<tag>!")
	( OX goes to !X!).  Case is not significant in tags,
	so OFOO and Ofoo will both find both !FOO! and !foo!.
	Label must be on same iteration	level as O command,
	i.e., no unmatched < or > between O and !.
	The label must also be within the same macro as
	the O;  in other words, non-local gotos are not
	implemented.
	If the tag is not found, an "UGT" error occurs at
	the end of the O command.  For convenience's sake,
	:O is just like O but simply returns if the tag
	is not found.
	The @ modifier allows the tag to be abbreviated.
	OFOO will not find !FOOBAR!, but @OFOO will find it.
	This is for the sake of those using O to do command
	dispatching.
	TECO has a cache containing the locations of
	several recent O commands and where they jumped to.
	If an O command is in that table, searching is
	unnecessary.  This increases efficiency.  However, if
	there is a ^] call in the arg to O, it might be
	intended to jump to different places each time, so
	TECO refuses to cache such jumps to force itself to
	search each time.  Also, jumps in buffers and in
	top level command strings cnnot be cached, since the
	data in the buffer (including the argument of the O
	command) might change at any time;  if TECO then did
	not read the argument and search, it might jump to the
	wrong place.

P	output contents of buffer to open output file, followed
	by form feed (^L) if FS ^LINSERT is 0; clear buffer and read
	into buffer from file open for reading until next form feed or
	end of file.  A single argument means do this that many times.

	The P command usually writes zero into
	the low bit of each word of the output file.  If the @ flag is
	set, then outputting on word boundaries in the buffer and the
	output file outputs the low bits as they are present in the
	buffer.

<m>,<n>P
	output the characters in the range <m> through <n> to the
	output file.  Do not change the buffer or read from the input
	file.

PW	outputs like P but does not
	clear buffer or do input.  Takes
	arg, meaning do it arg times.

Q<q>	returns the value in q-reg <q>, as a number.  If <q>
	is holding a number, that number is the value.
	If <q> "holds text", then it really contains a pointer
	to a string or buffer, and Q<q> wil return the pointer,
	which if put in another q-reg (using "U") that q-reg
	will "hold the same text" as <q>.

	A q-reg name is either an alphanumeric char preceded
	by 0, 1 or 2 periods,
	a "variable name" of the form <name>,
	a subscripting expression such as :Q(<idx>),
	a * (for certain commands),
	an expression in parentheses (for certain commands),
	or up to 3 periods followed by a "^R" or "^^" and any
	ASCII character.

	Periods plus alphanumeric character q-reg names refer
	to TECO's q-registers, which are what serve as variables
	for TECO programs.  Each distinct such name names a
	distinct variable.  Names with two periods are
	reserved for special system meanings;  those that
	are now assigned are documented starting at "..A".

	While names like "A" or ".8" are fine for local variables
	in programs, for global parameters mnemonic names are
	necessary.  Variables with long names are accessible through
	the <name> construct.  Variable names may be abbreviated,
	and extra spaces and tabs may go at the beginning, the end,
	or next to any space or tab.  Also, case is not
	significant inside variable names.  Thus, a variable named
	"Foo" could be accessed with QFOO, Q foo  or (if there
	is no FO or FOX, etc.) with Q Fo.  Because of this
	latitude, variables are not created if they are referenced
	and do not exist;  instead, they must be entered explicitly
	in the symbol table by the user.  This is easy to do,
	because the entire symbol table data structure is
	user-accessible; it is an FO-style symbol table that lives
	in q-register ..Q.  See the FO command for a sample macro
	for creating variables.  The symbol table must have at least
	two words per entry, but it may have more.  The additional
	words can be used however you like.  The FS VARMAC feature
	allows the third word to be used to hold a macro to be run
	when the value of the variable is changed, but this is enabled
	only if FS VAR MACRO is set nonzero.  See its definition.

	The elements of a q-vector may be accessed as q-registers
	in their own right.  If q-reg A contains a q-vector, then
	the "q-register" :A(0) is the first element of it, and
	:A(1) is the second, etc.  Indexing starts at zero for the
	first element of the q-vector, but only those elements
	within the virtual boundaries of the q-vector may be accessed.
	See FS QVECTOR for how to create q-vectors.

	A star ("*") may be used only with commands like ] and X
	that wish only to store in a q-register;  it causes such
	commands to return their data as a value instead.  Thus,
	:I*FOO returns a string containing FOO.

	Expressions in parentheses can be used only with
	commands that wish only to examine the contents of a
	q-register;  the value of the expression is used as
	the contents to be examined.  Commands that allow
	this option include F^A, F^E, F=, FQ, F~, G and M.
	Thus, G(Q0) is equivalent to plain G0.

	Q-reg names containing ^R or ^^ refer to the definitions
	of ^R-mode command characters.  When "^R" is used,
	the ^R-mode definition of the specified ASCII
	character is referred to;  when "^^" is used, the
	^R mode definition of the specified charcter xor'ed
	with 100 (octal) is meant.  The periods specify the
	control and meta bits since ^R-mode definitions
	belong to 9-bit characters but only 7-bit characters
	can be inside TECO command strings;  one period sets the
	control bit;  2, the meta bit;  3, both control and meta.
	If the char is obtained from a ^]^V, then all 9 bits
	may be obtained from that source;  the periods xor into
	the number in the q-reg.
	For example, "Q^RA" refers to the definition of "A",
	and "Q.^RA" refers to that of control-A, as does
	"Q^R^]^VX" when QX holds 301 (octal).
	"Q^R^A" refers to the definition of downarrow,
	one of the new TV printing characters, as does "Q^^A".
	"Q^^J" refers to the definition of linefeed, whereas
	"Q.^RJ" refers to the definition of control-J, which
	can be typed in only on a TV (and which is usually
	defined to execute the definition of linefeed).

R	move pointer left arg chars (no arg, same as arg=1).

:R	is to R as :S is to S; as :C is to C.
<m>,<n>r
	does "<m>+<n>-.J".  This is for FLR to work.

S	search.  Takes following text string
	and looks for it in the buffer,
	starting from the pointer.
	(if the string arg is null, the last nonnull arg
	to any search command is used)
	if it finds it, it positions the
	pointer after the string.  If it
	does not find it, it does not
	move the pointer but generates an
	error message unless the search is
	inside an iteration (see <.  See also FS S ERROR
	which may be used to disable this "feature").
	If the search is inside an iteration,
	the value as if produced by :s (read on)
	will be saved whether or not the
	colon is used, for use by the ;
	command.  The effect of iterations on searches is
	cancelled by errsets, so what matters is whether
	the search is more closely contained in an iteration
	or in an errset.
	A positive arg to the search means do it arg times,
	i.e., find the argth appearance of the string;
	a negative arg means search the buffer
	backwards from the pointer and
	position the pointer to the left of
	the string if successful.  If the S is
	preceded by "@", the char after
	the S is used to delimit the text
	string instead of altmode.  In this case, a null
	arg causes a search for the null string, instead of
	a search for the last string searched for.
	(this for the sake of macro-writers using ^])
	if the s is preceded by :, val=-1 if the
	search is successful and val=0 if
	not--there is no error condition.
	Note also the FB, N and _ commands, and the ^B command.

	There are some special characters which, when used inside
	search strings, do not have their normal meanings unless
	quoted with a ^Q:
	^X matches any character.
	^B matches any delimiter char (normally this means it
	matches any non-squoze char, but see q-reg ..D).
	^S<ch> matches any character whose Lisp syntax is <ch>.
	The Lisp syntax of a character is stored in q-reg ..D.
	^N matches any char other than the char that follows it in
	the search string.  ^N^B matches non-delimiters, and
	^N^X matches nothing.  ^N^Q^B matches all but ^B.
	^N^S( matches anything whose Lisp syntax is not "(".
	^O divides the string into substrings searched for
	simultaneously.  Any one of these strings, if found, satisfies
	the search.  Thus, SFOO^OBAR looks for FOO or BAR.
	If ^O is used inside a :S, finding the <n>th substring makes
	the search return -<n> as its value.
	^Q quotes the following char, i.e.,
	deprives it of special significance.
	In addition, ^] is special as usual.
	Note that SFOO^O will always succeed, and will move point
	forward over the next three characters
	if and only if they are FOO.  -2-(:SFOO^O) will do that and
	also return nonzero only if they were FOO.

T	type:  takes one or two args like K
	and types out the selected chars.
@T	types in the echo area.

U:
<n>U<q>
	puts the number <n> in q register <q>.
	Returns no value.
<m>,<n>U<q>
	puts <n> in <q>, and returns <m>.
	Thus "<m>,<n> U<q1> U<q2>" does "<n>U<q1> <m>U<q2>".

V	takes arg like K and displays chars, representing
	the cursor by "/\" (or whatever is in ..A).
	When, after being proceded from a --MORE--,
	a new screenfull is started, the place it began
	is remembered in FS WINDOW so that an attempt
	to display the buffer will try to start at the same place.
	This may make redisplay unnecessary if you search for
	something that appears on the screen.
	Nothing is typed on printing terminals.

@V	performs standard buffer display.
	That is, "@V" always does what automatic buffer
	display does as a default (when ..B holds 0).
	When in ^R mode, @V does a ^R-style display.
	Note that @V will display on any type of terminal,
	although TECO does not normally display automatically
	on printing terminals.
	In ^R mode, @V treats its arguments as ^R does
	(as hints on how to redisplay).  When not inside a ^R,
	@V ignores its arguments.

	If V is followed by W it becomes
VW	which does V, then waits for terminal
	input of one char whose 7-bit ASCII
	value is returned as val.

W	flushes current value except when
	part of VW or PW.

X	takes one or two args like K and
	enters selected chars as text into
	q-register named by next char in
	command string.  Can be retrieved
	by G command and ^] substitution, q.v.
	@X acts like X but
	appends text to q-reg rather than
	replacing q-reg contents.  If q-reg
	does not already contain text this
	works like ordinary X.
	See also :I.

Y	kills the buffer, then inserts one page from the
	current input file (until first formfeed or eof).
	Point is left at the beginning of the buffer.
	If reading is terminated by a ^L, the ^L will go in
	the buffer iff FS ^LINSERT is nonzero.
	(FS ^LINSERT is initially 0)
	The input file is not closed, even if eof is reached.
	To close the input file, use EC.  However, EE does
	close the input file.  Closing the input file is not
	necessary but will lighten the drain on system resources.
	Trailing ^C's or ^@'s just before eof are considered
	padding and are flushed.  To do input without having
	any padding characters removed, use FY.
	The virtual buffer boundaries are understood.
	If no file is open, the buffer is left empty.
	Because Y is an easy command to be screwed by, and
	isn't really necessary since the A command exists,
	there is a way to disable it.  See FS YDISABLE.

@Y	yanks in all the rest of the file.
	^L's within the file go in the buffer.
	A ^L at the end of the file will go in the buffer
	iff FS ^LINSERT is nonzero.  Trailing ^C's and ^@'s
	are conidered padding, and flushed.
	The input file is closed automatically.

Z	val=number of chars in buffer (more generally, the
	character number of the virtual end of the buffer,
	if virtual buffer boundaries are in use).
	See also FS Z, the actual length of the buffer,
	and FS BOUND and FS VZ, which set the virtual bounds.

[<q>	push text or number in q-reg <q> onto the q-register pdl.
	There is only one q-reg pdl, not one per q-reg.
	At various times (for example, the ^\ and F; commands,
	and after errors) TECO unwinds the q-reg pdl to a
	previous level by popping everything back to where it was
	pushed from.
	The [ command does not allow subscript expressions (such
	as "[:A(5)") because automatic unwinding would have no way
	to know how to pop the pushed value back where it came from.
	If you wish to push the value and don't mind that errors,
	etc. won't pop it back, do "[(Q:A(5))" or something similar.

<new>[<q>
	is equivalent to [<q> <new>U<q>.

\	Converts digits in the buffer to a number, or vice versa.
	If no arg, value is the number represented as decimal digits
	to right of pointer in buffer (actually, the input radix
	comes from FS IBASE, as with numbers in commands). 
	:\ allows an optional minus sign.  
	Moves pointer to right of number.

<n>\	inserts printed representation of <n> in
	buffer to right of pointer;  usually the number is
	"printed" in decimal, but the radix is controlled
	by q-reg ..E.
	If two args, first specifies field size such that if
	2nd is shorter than that many chars
	leading blanks will be added.
	Sets FS INSLEN to the number of characters inserted.
	See "FK" and "^F".
<m>,<n>\
	is like <n>\ but pads with leading spaces to <m> columns.  
<n>:\	returns a string containing the printed representation of
	<n>, in the radix in ..E.  Sets FS INSLEN.
<m>,<n>:\
	makes a <m> character string, with leading spaces.

]<q>	pop from q-register pdl into q-reg <q>.
	See the [ command.

^	This command has been replaced by the @ command
	but still works.  See the @ command.

_	if FS _ DISABLE  is 0, then _ is like S,
	except at end of buffer do Y command and continue search
	until end of file on input or until text string found.
	If FS _ DISABLE is 1, "_" is illegal.
	If FS _DISABLE is -1, "_" is the same as "-".
	Use "F_" in a macro to be sure of doing the search.

Rubout	deletes last char typed in, and types
	deleted char.  Done during type-in,
	not during command execution.
	If executed (rather hard to do), same as _.
	Rubouts are typed out by TECO as ^? (rubout is ctl-?)

	Lower case letters are interpreted like
	upper case letters when they are commands.
	Inside insert and search strings they are
	treated as themselves.

Various special topics of interest are treated below

When TECO is started for the first time,
	it initializes various data areas, prints its
	version number, and initializes several flags associated
	with the terminal (by executing FS TTY INIT).
	On T(w)enex, if TECO was started at 1 + the normal starting
	address, it expects to be passed a JCL and saves the JFN's
	filenames, as a string, in FS CCL FNAME.
	If TECO was started at 2 + the normal starting
	address, FS LISPT is set nonzero.  Otherwise, it is set to 0.
	In either case, TECO looks for a "TECO INIT" file (see below),
	executing it as a program if it is found.

When TECO is restarted,
	it does not clobber the buffer, q-regs or open files.
	It does, however, execute FS TTY INIT which resets some
	flags whose preferred setting depend on the type of terminal.
	FS TTY INIT calls the user-supplied FS TTY MACRO, if not 0.
	Then, it quits to top level and executes
	whatever is in q-reg ..L (unless it is 0).

Init files:
	whenever TECO is started for the first time, it checks
	for a file <hsname>;<xuname> TECO, for a file <hsname>;* TECO,
	and then for the file .TECO.;* TECO.  (On T(w)enex, it looks
	for <conn-dir>TECO.INIT and then for EMACS:TECO.INIT).
	The first one found is
	executed as TECO commands.  The last of those files is the
	default init file.  The other two would be personal init files.
	The personal init file can do whatever you want.
	The only thing the default init file does is interpret command
	lines from DDT as follows:

	":TECO FOO BAR <cr>"     typed at DDT causes    
		"ET FOO BAR  EI ER Y" to be done by TECO --
		that is, TECO starts editing FOO BAR.
	":TECO FOO <cr>" edits FOO >  .
		Because COM:.TECO. (INIT) sets FS FNAMSY temporarily.
	":TECO <filename><TECO commands> <cr>"
		typed at DDT executes " ET <file>  <commands> ".

	A user's own init file should interpreted the JCL by
	reading it in with the "FJ" command.  It may have any
	command format it wishes.

	Dumped TECO programs runnable as separateprograms often have
	their own init files.  Such packages should use FS XUNAME and
	FS HSNAME to determine init file names the same way TECO
	does.


Journal Files

	TECO can write a journal file describing all the commands
	typed by the user.  If the system crashes, the user can replay
	the journal file later so that his work is not lost.

	The TECO program starts writing a journal file by doing FS JRN
	OPEN.  This opens the file for output, using the default
	filenames.  TECO automatically writes into the file
	representations of all command characters read from the
	terminal.  To close the journal file, use :FS JRN OPEN.  To
	see whether a journal file is being written, look at the value
	of FS JRN OUT (it is -1 in that case).  Note that no output
	is ever written in a journal file while another journal file
	is being replayed.

	The TECO program requests the replaying of a journal file by
	doing FS JRN EXECUTE.  This opens the file using the default
	filenames.  When TECO is ready to read more commands, they
	will be read from the journal file until the file is
	exhausted, or the user types a ^G on the terminal, or the TECO
	program does :FS JRN EXECUTE to close the file.  FS JRN IN
	is -1 when a journal file is being replayed.

	Each command character is represented by a pair of characters
	in the journal file.  This is so that the Control and Meta
	bits can be represented in full generality.  A character with
	neither the Control nor the Meta bit is represented in the
	journal file by a space followed by the character itself.  A
	Control character (200+nnn) is represented by an "^" followed
	by the non-control character.  A Meta character (400+nnn) is
	represented by "+" followed by the non-meta character.  A
	Control-Meta character is represented by "*" followed by the
	basic character.  Thus, Meta-F is represented by "+F".

	There are a few exceptions to the rule above.  A CR character
	(015) is represented by the two characters CRLF.  This makes
	the journal file look much nicer.  The Help character is
	represented by "??".

	Other constructs which can appear in a journal file are ^G
	(007) to represent a quit command; comments, which start with
	semicolon and end with CRLF, and are ignored when replaying
	the file; and program commands, which start with a colon and
	whose format is up to the TECO program.  When a ^G or a colon
	is seen in a journal file being replayed, the FS JRN MACRO is
	called, and given the character (^G or colon) as an argument.
	In the case of ^G, the macro should do 1F[JRN INHIBIT and
	then call ^R so that the user
	can fix up for any timing errors that occurred in quitting,
	and when the ^R returns it should do -1FS QUIT to replay the
	quit.  Colon represents the start of program-specific
	information; the program which wrote the journal file and the
	one which replays it must agree on what format the following
	information takes.  The information can be read using the FS
	JRN READ command.

	One specific colon-sequence is defined by TECO:
	colon followed by a ^G indicates a quit which happened
	synchronously, while TECO was waiting for input.  Such a quit
	does not need a recursive ^R for the user to fix things up
	because everything is guaranteed correct.  In this case, the
	FS JRN MACRO should just replay the quit with -1FS QUIT.

	Normally, journal files are written and read automatically.
	However, the TECO program may wish to add extra information to
	the journal file.  This could be comments designed to aid
	humans who look at the journal file, or it could be
	information intended to be used in replaying the file
	properly.  In that case, it should take the form of a colon
	followed by information to be read in by FS JRN MACRO later.
	In either case, the information is written by means of the FS
	JRN WRITE command, which accepts a character or a string.


TECO's Data Structures (Strings, Buffers and Qvectors)

	  TECO has two different data structures for storing sequences
	of characters: strings, and buffers.  They differ in what
	operations are allowed on them, and how efficient they are.

	  Strings have less overhead than buffers, but as a penalty
	they are not easily altered.  Once a string has been created,
	its contents usually do not change; instead one might
	eventually discard the string and create a new one with
	updated contents.  The sole exception is F^E, which makes it
	possible to alter characters in a string (but not to insert or
	delete).  Commands which "put text in a q-register" all do so
	by consing up a string and putting a pointer to it in the
	q-register.

	  Buffers are designed to be convenient for insertion and
	deletion.  Each buffer has its own pointer, and its own
	virtual buffer boundaries, which are always associated with
	it.  The contents of a buffer can be accessed just like the
	contents of a string (in which case only the part between the
	virtual boundaries is visible), but it can also be "selected"
	and then accessed or altered in many other ways: insertion,
	deletion, searching, etc.  Each buffer also has a flag which
	is set nonzero whenever the buffer's contents are changed; it
	can also be set by the user with FS MODIFIED.  A buffer has
	about 42 characters of overhead, and the number of buffers is
	limited (about 40).  Initially, there is only one buffer in a
	TECO (pointers to which are initially found in q-registers ..O
	and ..Z), and new ones are only made when explicitly asked for
	with F[ B BIND, FS B CONS or FS B CREATE.

	  Strings and buffers are normally represented in TECO by
	pointers.  When a q-register "contains" a string, it actually
	contains a pointer to the string (see the sections on internal
	format for details).  If q-register A contains a string, QA
	returns the pointer, which can be stored into q-register B;
	then QB and QA both point to the same string.  The command
	:I*<string> returns a pointer to a newly consed up string
	containing <string>.  :G can be used to extract a substring
	(which is copied, not shared).  The commands :\, :F6 and X*
	are also useful for creating strings.

	  A buffer is selected by putting a copy of a pointer to it
	into q-register ..O.  TECO has a garbage collector, so that if
	all pointers to a buffer or string are eliminated, the storage
	it occupies will eventually be reclaimed.  Most of the space
	occupied by a buffer can be reclaimed explicitly with the FS B
	KILL command; the buffer is becomes "dead", and even though
	pointers to it may still exist, any attempt to use them to
	select the buffer or examine its contents will be an error.

	  Vectors of objects can also be represented in TECO, with
	either buffers or qvectors.  Buffers can be used to as vectors
	of numbers, while qvectors are used as vectors of arbitrary
	objects (numbers, or pointers to strings, buffers or
	qvectors).  The difference is due to the fact that the garbage
	collector knows that the objects in a qvector might be
	pointers and therefore must be marked, while the objects in a
	buffer cannot be pointers and are ignored.  The words in a
	buffer or q-vector can be accessed easily with subscripted
	q-register names; if QA contains a q-vector, then Q:A(0) is
	its first element.  To access the elements in hairier ways,
	you can select the buffer or q-vector and the insert or
	delete, etc.  Q-vectors are created by means of FS Q VECTOR.



The buffer block, and what buffers contain (and the gap):

	The current buffer is described by the 7-word "buffer block"
	which contains these variables:

	 BEG	char addr of start of buffer,
	 BEGV	char addr of lower buffer boundary,
	 PT	char addr of pointer,
	 GPT	char addr of start of gap,
	 ZV	char addr of upper buffer boundary,
	 Z	char addr of top of buffer,
	 EXTRAC	# chars in gap.
	(next come 2 words used by the communication
	 protocol.  See below).
	 MODIFF	nonzero if buffer has been modified.
	 RDONLY value of FS READ ONLY for this buffer.

	Note that all character addresses normally used in TECO have
	BEG subtracted from them; "B" returns BEGV-BEG; "Z", ZV-BEG;
	"FS Z", Z-BEG; ".", PT-BEG; "FS GAP LOCATION", GPT-BEG.  "FS
	GAP LENGTH" gives EXTRAC.  The actual value of BEG is
	available as "FS REAL ADDRESS".

	  GPT and EXTRAC describe the "gap", a block of unused space
	in the middle of the buffer.  The real amount of space used by
	the buffer is Z-BEG+EXTRAC.  BEGV, PT, Z and ZV are "virtual"
	addresses in that they take no account of the gap.  To convert
	a virtual address to a real one, add EXTRAC to it if it is
	greater than or equal to GPT.  Real address 0 refers to the
	first character in word 0; real address 5 refers to the first
	character in word 1, etc.

	  It is OK for the superior to alter those variables or the
	contents of the buffer, if TECO is between commands or has
	returned because of ^K, FS EXIT or ^C; except that BEG should
	not be changed and the sum of Z and EXTRAC (the real address
	of the end of the buffer) should not be changed, unless
	appropriate relocation of other buffers and TECO variables is
	undertaken.  TECO programs need not worry about the gap,
	except for efficiency reasons or when communicating with
	machine language programs, and they need never convert
	character addresses to real addresses; TECO does all that.


Strings - internal format:

	  A string containing <n> characters takes up <n>+4
	consecutive characters in TECO.  It need not start on a word
	boundary.  The first four characters are the string header;
	the rest, the text of the string.  The header starts with a
	rubout.  The second character is <n>&177; the third,
	(<n>/200)&177; the fourth, <n>/40000 (numbers in octal).


Buffers - internal format:

	  A buffer consists of a buffer-string, which points to
	a buffer frame, which points to the buffer's text.

	  The buffer-string is similar to a string, and exists in a
	string storage space, but begins with a "~" (ASCII 176)
	instead of a rubout.  It has only three more characers; the
	second is <addr>&177; the third, (<addr>/200)&177; the fourth,
	<addr>/40000; <addr> being the address of the buffer frame.

	  The buffer frame is a seven-word block whose purpose is to
	save the buffer block for buffers which are not selected.
	While a buffer is selected, the buffer frame contents may not
	be up to date.

	  The first word of the frame contains a few flag bits in the
	left half.  The sign bit will be set to indicate that the
	block is in use as a buffer frame.  The 200000 bit is the GC
	mark bit and should be ignored.  The 100000 bit, if set,
	indicates that the buffer is really a qvector.  The 40000 bit
	is the "buffer modified" flag.  These bits are only in the
	buffer frame, not the buffer block (BEG).


Buffer and string pointers - internal format: 

	  When a q-reg is said to hold a buffer or a string, it really
	contains a pointer to the buffer or string.  The pointer is in
	fact a number, distinguished from other numbers by its value
	only!  A range of the smallest negative numbers are considered
	to be pointers (this is why QAUB copies a string pointer from
	QA to QB without any special hair).

	  Numbers which are pointers are decoded by subtracting
	400000000000 octal (the smallest negative integer) to get a
	character address.  This may either be the exact address of a
	character in pure space (what :EJ loads into), or the relative
	address of a character in impure string space (what "X"
	allocates within.  The char address of the start of impure
	string space is held in location QRBUF).  In either case, that
	character should be the rubout beginning a string or the "~"
	starting a buffer-string.  If the number, thus decoded, does
	not point within either of those ranges, or doesn't point at a
	"~" or rubout, then it is not a pointer - just a plain number.

	  For example, 400000000000.+(FS :EJPAGE*5*2000.) is a string
	pointer to a string whose first character is at the very
	beginning of the last :EJ'd file.  If the file has the proper
	format (see "strings" above), that number may be put in a
	q-reg and the string then executed with "M" or gotten by "G",
	etc.  The file might contain a buffer-string except that
	causing it to point to a legitimate buffer frame would be
	difficult.  Making it point to a counterfeit buffer frame
	inside the file would lose, since TECO tries to write in
	buffer frames.


Low Bits:

	   On ITS, the low bits of the words of a "text" file are
	generally ignored, but on other systems such as BOTTOMS-10 any
	word whose low bit is 1 is effectively not present in the
	file.  Therefore, it is desirable not to set low bits of text
	files.  It is also necessary to be able to manipulate binary
	files, reading and writing all the bits.

	TECO never touches the low bits of files when they are read
	in.  When a file is written out, normally all the low bits are
	cleared.  However, the @P command outputs a file without
	touching the low bits.  It is used to output binary files.

	When operating on binary files data in the buffer, you must be
	careful to avoid operations which can clear the low bits of
	the words.  Anything which moves text except on word
	boundaries can shuffle the positions of the characters in the
	words, and the low bits will be misplaced if not also
	cleared.  You can avoid such problems if you never insert or
	delete characters except in groups of 5 or multiples of 5,
	starting at positions which are multiples of 5.  In this
	context, note that 5,0I will NOT clear the low bit; you must
	do 0,.-5FS WORD or 0U:..Q(<n>) (or something else suitable)
	to do that.

How superiors can put text (and other things) into TECO:

	   A standard protocol for communication from a superior to
	TECO is hereby defined, which allows the superior to request
	space in the buffer for inserting text, or request that a file
	be loaded and a certain function be found.  Macro packages may
	supply functions to handle the requests instead of TECO's
	default handler.

	A superior can make a request whenever TECO has deliberately
	returned control (by a ^K valret, FS EXIT or a ^C) by
	restarting TECO at a special place: 7 plus the address of the
	"buffer block", which address can be found in accumulator 2 at
	such times.  Save the old PC before setting it, since you must
	restore the PC after the request is handled.  The word after
	the starting location (8 plus the buffer block address) is
	used for an argument.

	There are two types of requests.  If you wish to supply text
	for TECO to edit, the argument should be the number of
	characters of text you need space for (it may be 0).  In that
	case, TECO will return (with an FS EXIT) after making at
	least that much space in the buffer, after which you should
	restore the PC at which TECO had stopped before you made the
	request.  You can then insert the text in the buffer and
	restart TECO.

	If you want TECO to read in a file, supply a negative
	argument and be prepared to supply TECO with JCL, when it asks
	with the standard .BREAK 12, describing what it should do.
	When TECO does a .BREAK 16, (FS EXIT) you can assume it has
	done its work, and should restore the old PC.  The formats for
	the JCL string are <filename><cr>, <filename>,<decimal
	number><cr>, and <filename>,<function name><cr>.  A decimal
	number should be the address within TECO of the place to start
	editing.  A function name can be anything that isn't a number,
	and its interpretation is not specified.

	TECO macro packages can supply a handler for requests from
	the superior by putting it in FS SUPERIOR.  It will receive
	the argument describing the type of request as its numeric
	argument (^Y), and can read the JCL with FJ and do an FS EXIT
	when finished.  If FS SUPERIOR is zero, TECO's default
	actions will be taken.  Note that TECO's default handling of a
	request to load a file is to do nothing.


TECO's character sets:
   (numbers in this section are in octal)

	The most important TECO character sets are ASCII (7-bit)
	and the 9-bit TV set.  The contents of all files, strings,
	and buffers, and thus all TECO commands, are in ASCII;
	9-bit is used only for terminal input.  Here is how TECO
	converts between character sets:

	14-bit to 9-bit conversion when characters are read in:

	When a character is actually read from the terminal, it is in
	a 14-bit character set which contains a basic 7-bit code,
	and the control, meta and top bits (also shift and shift-
	lock, which are ignored since they are already merged into
	the basic 7-bit character).  TECO converts it to 9-bit
	as follows:  if top is 0, and the 7-bit character
	is less than 40 and not bs, tab, LF, CR or altmode,
	then add control+100; then clear out top, shift and shift-lock.
	Thus, TV uparrow comes in as top+013 and turns into 013;
	TV control-K comes in as control+113 and stays control+113;
	TV "VT" comes in as 013 and turns into control+113;
	TV control-VT comes in as control+013 and becomes control+113;
	non-TV control-K comes in as 013 and becomes control+113;
	TV control-I comes in as control+111 and stays control+111;
	TV "tab" comes in as 011 and stays 011;
	TV control-tab comes in as control+011 and stays control+011;
	non-TV "tab" or control-I comes in as 011 and stays 011.

	The character 4110 (Top-H) is the HELP character;  it is
	handled specially.  It is not changed, even when any other
	character would be converted to 9-bit or 7-bit.  It is always
	returned as 4110.  However, if FS HELP MAC is nonzero, then
	4110 typed AT ANY TIME will cause FS HELP MAC to be run.
	When FS HELP MAC is run, the 4110 is otherwise ignored, so
	TECO will continue to wait for input (unless FS HELP MAC
	sets FS REREAD to supply some).  Only when FS HELP MAC is
	zero will 4110 ever be returned to the user.  ^R ignores
	4110 (except for running FS HELP MAC).

	9-bit to ASCII, when TECO wants to read an ASCII code:

	input read in using "@FI", or read by the ^R-mode
	comand dispatch, is used as 9-bit.  However,
	when input is read by "FI", or by the "^T" command reader,
	or by TECO top level, it must be converted to ASCII
	as follows:  meta is thrown away;  if control is 0 then
	nothing changes;  otherwise, control is cleared and the
	following actions performed on the 7-bit character that
	is left:  rubout stays the same;  characters less than 40
	stay the same;  characters more than 137 have 140 subtracted;
	characters 100 to 137 have 100 subtracted;  all others
	are unchanged, except for 40 (Space) which becomes 0 (^@).
	Thus, control+111 (TV control-I) becomes 011;
	control+011 (TV control-tab) becomes 011;
	and 011 (TV tab, or non-TV control-I) stays 011.
	Similarly, TV uparrow, TV "VT", TV control-K and non-TV
	control-K all become 013.

ASCII to 9-bit in FS ^RCMACRO and FS ^R INIT:

	The ^R command dispatch table is indexed by 9-bit
	characters.  For compatibility with the time that it was
	not, the commands FS ^R CMACRO and FS ^R INIT, when
	not given the atsign modifier, accept an ASCII argument,
	and try to have the effect of referring to the definition
	of that ASCII character - in fact, they convert the ASCII
	character to 9-bit and then index their tables.  The
	conversion is as follows:
	if the character is less than 40, and is not bs, tab,
	LF, CR or altmode, then add control+100.
	Thus, 013 (^K) becomes control+113 (TV "VT" or control-K,
	not TV "uparrow"), which is just right.  Tab, etc. have
	a harder time doing the right thing, since both 011
	control+111 are plausible ways that the user could type
	what corresponds to ASCII 011.  The solution chosen is
	to leave 011 ASCII the same in 9-bit, since the ^R-mode
	definition of control-111 is to use 011's definition.

The initial ^R-mode definitions of all 9-bit characters:

	 All characters whose bottom 7 bits form a lower case
	letter are defined to indirect through the corresponding
	upper case character.  Their definitions are all
	40,,RRINDR, where RRINDR is the indirect-definition routine,
	and 40 specifies the character 40 less.
	 Control-BS and Control-H indirect through BS, and similarly
	for Tab and LF.  Control-CR and Control-Altmode (but not
	Control-M and Control-[) indirect through CR and Altmode.
	An isomorphic indirection-pattern exists for meta characters.
	 Control-, Meta- and Control-Meta-digits all accumulate into
	the numeric argument.
	 Control-, Meta- and Control-Meta-Minus-sign all set the
	4 bit in FS ^R ARGP, thus negating the next command's arg.
	 All other meta characters are self-inserting.  A few
	(mentioned above) are self inserting because they go indirect
	through other meta characters.
	 All non-control non-meta characters, except for CR,
	altmode and rubout, are self-inserting.  CR inserts CRLF;
	altmode leaves ^R-mode;  rubout deletes backwards.
	Of the rest, ^H, ^I and ^J are defined to insert themeselves
	straight away, while the rest are defined to be "normal"
	and do whatever FS ^R NORMAL and FS ^R REPLACE say.
	 Control-rubout has its own special routine, which deletes
	treating spaces as if they were tabs.
	 Control-digits update the numeric arg for the next command.
	 All other control characters not in the range
	control+101 through control+135 are errors.
	 Control-M inserts just a CR.  Control-[ is an error.
	 The remaining control characters from control-101 to
	control-135 do what the ^R command table says, or else are
	errors.


Neat hacks that are frequently useful:

1) To take an absolue value, use ^@^@.  -1^@^@ and 1^@^@ are both 1.
  This assumes that there is no pre-comma argument.
2) When there is just a pre-comma argument, :^@ will make it a
  post-comma argument instead.  There will be zero as a pre-comma
  argument, but another ^@ will flush that.  Thus, :^@^@ gets a
  pre-comma argument as a post-comma argument with no pre-comma.
  If there is a post-comma argument already, start with *0 to clear it
  first, giving *0:^@^@.
  Since many commands ignore pre-comma arguments, simply :^@ may be
  enough.
3) ",0^@" negates its argument, like "*(-1)".
4) The easiest way to compute the maximum of two quantities is with
  F^@, as in QA,QBF^@.  This, however, returns the other one as a
  pre-comma argument.  For the minimum as a post-comma arg, add :^@.


TECO program writing standards:

1) Each line that doesn't begin inside a string argument
   should be indented with at least one space,
   preferably with a number of spaces
   indicating how deep in conditionals and
   iterations it is.
2) The semantic content of one line of TECO program
   should be no greater than that of one line of any
   other language, if the program is to be understandable.
   In other words, break lines frequently - and put a
   comment on each line.  There should be spaces
   between logical groups of commands, every few characters,
   as in "3K J IFOO", which also shows how long a line
   should be.
3) The standard way to write a comment is to make it
   look like a tag:  !<comment>!.
4) Follow value-returning commands with "W"'s or CRLFs
   when the value is not used for anything.
5) An example of a well-commented TECO program is RMAIL.
   See EMACS1;RMAILX > and RMAILZ >.  Note that these are
   written in EMACS compressor format.


TECO program debugging aids:

1) Trace mode causes all TECO commands to be typed
   out as they are executed.  See the "?" command.
   A good technique is " 1F[TRACE MA " for running
   q-reg A in trace mode.
2) Using FS STEP MACRO you can cause TECO to pause at each line,
   displaying the buffer and waiting for input before
   continuing execution.  This works best when lines
   are short, as they ought to be anyway.
3) Break-loops on errors are available, by setting
   FS *RSET to nonzero, or by using Q..P.
4) It is easy to edit a "FTHere I am" or "Q0=" into
   the program and re-execute it.
5) If the standard top level is in use, "?" typed in
   after an error will cause a printout of a short
   section of command string up to the point of the error.
   User break loops in Q..P can use FS BACKTRACE to do
   similar or better things.
6) Setting FS .CLRMODE to 1 disables the ^L and F+
   commands, which normally clear the screen.  This may
   be useful for debugging programs that wipe out their
   trace output.

Local Modes:
Mode:Text
Fill Column:70
Space Indent Flag:1
Paragraph Delimiter:
	:m.m ^R Indent Relative
End: