wt_config.xml.in

<!--
    Wt Configuration File.

    The Wt configuration file manages, for every Wt application, settings
    for session management, debugging, directory for runtime information
    such as session sockets, and some security settings.

    Settings may be specified globally, or for a single application path.

    The path should be as configured in the Wt build process, where it
    defaults to /etc/wt/wt_config.xml. It can be overridden in the environment
    variable WT_CONFIG_XML, or with the -c startup option of wthttpd.

    The values listed here are the default values, which are used when the
    declaration is missing or no configuration file is used.
  -->

<server>

    <!-- Default application settings

      The special location "*" always matches. See below for an example
      of settings specific to a single application.
      -->
    <application-settings location="*">

        <!-- Session management. -->
	<session-management>
            <!-- Every session runs within a dedicated process.

	       This mode guarantees kernel-level session privacy, but as every
	       session requires a seperate process, it is also an easy target
	       for DoS attacks if not shielded by access control.

               It is also a convenient mode during development, as it is easy
	       to enable disable debugging using valgrind, and it always starts
	       the latest deployed executable for a new session.
	   
	       Note: currently only supported using the FastCGI connector
              -->

	    <!--
	       <dedicated-process>
		 <max-num-sessions>100</max-num-sessions>
	       </dedicated-process>
	      -->

	    <!-- Multiple sessions within one process.

	       This mode spawns a number of processes, and sessions are
	       allocated randomly to one of these processes (you should not
	       use this for dynamic FCGI servers, but only in conjunction
	       with a fixed number of static FCGI servers.

	       This requires careful programming, as memory corruption in one
	       session will kill all of the sessions in the same process. You
	       should debug extensively using valgrind. Also, it is your
	       responsibility to keep session state not interfering and
	       seperated.

	       On the other hand, sessions are inexpensive, and this mode
	       suffers far less from DoS attacks than dedicated-process mode.
	       Use it for non-critical and well-debugged web applications.

	       Note: wthttpd always uses exactly one process
              -->
	    <shared-process>
	        <num-processes>1</num-processes>
	    </shared-process>

	    <!-- Session tracking strategy.

	       Possible values:
		 Auto: cookies is available, otherwise URL rewriting
		 URL:  only URL rewriting
	      -->
	    <tracking>URL</tracking>

	    <!-- How reload should be handled.

	       When reload should (or rather, may) spawn a new session, then
	       even in the case cookies are not used for session management,
	       the URL will not be cluttered with a sessionid.
	       However, WApplication::refresh() will never be called.
	      -->
	    <reload-is-new-session>true</reload-is-new-session>

	    <!-- Session timeout (seconds).

	       When a session remains inactive for this amount of time, it is
	       cleaned up.
	      -->
	    <timeout>600</timeout>

	    <!-- Server push timeout (seconds).

               When using server-initiated updates, the client uses
               long-polling requests. Proxies (including reverse
               proxies) are notorious for silently closing idle
               requests; the client therefore cancels the request
               periodically and issues a new one. This timeout sets
               the frequency.
	      -->
	    <server-push-timeout>50</server-push-timeout>
	</session-management>

	<!-- Settings that apply only to the FastCGI connector.

	   To configure the wthttpd connector, use command line options, or
	   configure default options in /etc/wt/wthttpd
	  -->
	<connector-fcgi>
	    <!-- Valgrind path

               If debugging is enabled and this path is not empty, then valgrind
	       will be started for every shared process, or for every session
	       which has ?debug appended to the command line.

	       The variable is slighly misnamed. Not only a path can be set,
	       but also options, like for example:

                 /usr/bin/valgrind - -leak-check=full
             -->
	    <valgrind-path></valgrind-path>

	    <!-- Run directory

               Path used by Wt to do session management.
	      -->
	    <run-directory>${RUNDIR}</run-directory>

	    <!-- Number of threads per process

               This configures the size of the thread pool. You may
               want to change this value if you would like to support
               reentrant event loops, where you block one event loop
               using WDialog::exec() or related static
               methods. Everytime you enter such an event loop, one
               thread is blocked, and therefore the total number of
               sessions that reliably can do this is limited to the
               number of thread you have (minus one to unblock).

	       For the built-in http connector, there is a similar
	       config option that is specified in the whttpd config
	       file or on the command line (-t).

               The default value is 1.
	      -->
	    <num-threads>1</num-threads>

	</connector-fcgi>

	<!-- Settings that apply only to the MS IIS ISAPI connector.

	   To configure the wthttpd connector, use command line options, or
	   configure default options in /etc/wt/wthttpd
	  -->
	<connector-isapi>
	    <!-- Number of threads per process

               This configures the number of threads that will be used
               to handle Wt requests. The IIS internal threads are never
               used to do any processing; all requests are forwarded to
               be handled in this threadpool. Rather than to configure a
               so-called 'web-garden' in IIS, increase this number. The
               ISAPI connector will not work correctly when a web-garden
               is configured.

               You may want to change this value if you would like to
               support more reentrant event loops, where you block one
               event loop using WDialog::exec() or related static
               methods. Everytime you enter such an event loop, one
               thread is blocked, and therefore the total number of
               sessions that reliably can do this is limited to the
               number of thread you have (minus one to unblock).

               You may also want to increase this number if your Wt
               application is regularly waiting for IO (databases, network,
               files, ...). If this number is too low, all threads could
               be waiting for IO operations to complete while your CPU
               is idle. Increasing the number of threads may help.

               Computing intensive applications may also increase this number,
               even though it is better to offload computations to a helper
               thread and user server push or a WTimer to check for
               completion of the task in order to keep your GUI responsive
               during the computations.

               The default value is 10.
	      -->
	    <num-threads>10</num-threads>

	    <!-- Maximum Request Size spooled in memory (Kb)
        
               Normally, Wt keeps incoming requests (POST data) in memory.
               However, malicious users could send a big POST and as such
               use up all memory of your HTTP server. With this parameter,
               you tune how big a request can be before Wt spools it in a
               file before processing it. Legitimate big POST messages may
               occur when users are expected to upload files.

               See also max-request-size.

               The default value is 128K, which is more than enough for
               any interactive Wt event.
	      -->
	    <max-memory-request-size>128</max-memory-request-size>
	</connector-isapi>

        <!-- Javascript debug options

	     Values:
	     - true: JavaScript errors are not caught but the browser
	             built-in debug options are used
	     - stack: JavaScript errors are caught but also a stack trace
                      is printed, useful for diagnosing a problem in production
	     - false: JavaScript errors are caught and a simple error message
	              is printed to indicate that something is terribly wrong
	  -->
	<debug>false</debug>

	<!-- Log file

	   When the log file is empty, or omitted, logging is done to
	   stderr. This may end up in the web server error log file
	   (e.g. for apache + fastcgi module), or on stderr (e.g. for
	   the built-in httpd).
         -->
	<log-file></log-file>

	<!-- Logger configuration

	   This configures the logger. With the default configuration,
	   everything except for debugging messages are logged.

	   The configuration is a string that defines rules for
	   enabling or disabling certain logging. It is a white-space
	   delimited list of rules, and each rule is of the form:

	     [-]level : enables (or disables) logging of messages of
	     the given level; '*' is a wild-card that matches all
	     levels

	     [-]level:scope : enables (or disables) logging of
	     messages of the given level and scope; '*' is a wild-card
	     that matches all levels or scopes.  The default
	     configuration is "* -debug", i.e. by default everything
	     is logged, except for "debug" messages.

	   Some other examples:

	     "* -debug debug:wthttp": logs everything, including
	     debugging messages of scope "wthttp", but no other
	     debugging messages.

	     "* -info -debug": disables logging of info messages
	     in addition to debugging messages.

	   Note debugging messages are only emitted when debugging
	   has been enabled while building Wt.
	  -->
	<log-config>* -debug</log-config>

	<!-- Maximum HTTP request size (Kb)

           Maximum size of an incoming POST request. This value must be
           increased when the user is allowed to upload files.
         -->
	<max-request-size>128</max-request-size>

	<!-- Session id length (number of characters) -->
	<session-id-length>16</session-id-length>

	<!-- DoS prevention: limit plain HTML sessions

           This is a simple measure which avoids Denial-of-Service
           attacks on plain HTML sessions, which are easy to mount in
           particular in the case of progressive bootstrap mode.

           This setting may be used to keep the ratio of plain HTML
           versus Ajax sessions under a certain ratio. Typically, most
           of your sessions will be Ajax sessions, and only a tiny
           fraction (e.g. less than 5%) will be plain HTML
           sessions. This ratio is only enforced when more than 20 sessions
	   have been created.
	  -->
	<plain-ajax-sessions-ratio-limit>1</plain-ajax-sessions-ratio-limit>

	<!-- DoS prevention: adds a puzzle to validate Ajax sessions

           This is a simple measure which avoids Denial-of-Service
           attacks on Ajax sessions.

           When enabled, a puzzle needs to be solved in the first Ajax
           request which eliminates agents that do not build a proper
           DOM tree.
	  -->
	<ajax-puzzle>false</ajax-puzzle>

	<!-- Send the XHTML mime type when appropriate

	   Wt renders XHTML1 (XML variant of HTML) that is backward-compatible
	   with HTML. Using XHTML, Wt is capable of supporting XHTML-only
	   features such as embedded SVG or MathML.

	   When enabled, JWt sets an XHTML mime-type
	   (application/xhtml+xml) when the browser reports support
	   for it. Most notably, Internet Explorer does not support
	   it.  Because XHTML and HTML are slightly different with
	   respect to default CSS rules, you may want to disable
	   sending the XHTML mime-type alltogether, at least if you
	   are not using SVG (used by the WPaintedWidget).  -->
	<send-xhtml-mime-type>false</send-xhtml-mime-type>

	<!-- Do strict serialization of events.

	   By default events are queued at the client-side, and
	   transmitted to the server so that at any time only one
	   request/response is pending. This scheme has a quality that
	   resembles TCP: on a low-latency link you allow the
	   transmission of many smaller requests, while on a high
	   latency link, events will be propagated less often, but in
	   batches.

	   In any case, this scheme does not drop events, no matter
	   how quickly they are generated.

	   In rare cases, the scheme may result in unwanted behaviour,
	   because the client-side is allowed to be slighly out of
	   sync at the time an event is recorded with the server-side
	   (and more so on high-latency links). The drastic
	   alternative is to discard events while a response is
	   pending, and can be configured by setting this option to
	   true.
	  -->
	<strict-event-serialization>false</strict-event-serialization>

	<!-- Enables web socket.

	   By default Ajax and long polling are used to communicate
	   between server and browser.

           By enabling web socket support, if the browser supports
           WebSockets, then WebSocket is the protocol used for
           communication between client and server. WebSockets are
           currently only supported by the built-in httpd Connector,
	   which acts as both an HTTP and WebSocket server. The WebSocket
	   protocol is intentionally not compatible with HTTP, through 
	   a special hand-shake mechanism, and requires
           that all (reverse) proxy servers also have explicit support
           for this protocol.

           This feature is still experimental: the Web Sockets RFC is
           still a draft. Wt implements up to version 17 of the draft
	   (latest as of November 2011).
	  -->
	<web-sockets>false</web-sockets>

	<!-- Redirect message shown for browsers without JavaScript support

	   By default, Wt will use an automatic redirect to start the
	   application when the browser does not support
	   JavaScript. However, browsers are not required to follow
	   the redirection, and in some situations (when using XHTML),
	   such automatic redirection is not supported.

	   This configures the text that is shown in the anchor which
	   the user may click to be redirected to a basic HTML version
	   of your application.
          -->
	<redirect-message>Load basic HTML</redirect-message>

	<!-- Whether we are sitting behind a reverse proxy 

	   When deployed behind a reverse proxy (such as Apache or Squid),
	   the server location is not read from the "Host" header,
	   but from the X-Forwarded-For header, if present.
	  -->
	<behind-reverse-proxy>false</behind-reverse-proxy>

	<!-- Whether inline CSS is allowed.

           Some Wt widgets will insert CSS rules in the the inline
           stylesheet when first used. This can be disabled using this
	   configuration option.

           Note: some widgets, such as WTreeView and WTableView,
           dynamically manipulate rules in this stylesheet, and will
           no longer work properly when inline-css is disabled.
	  -->
	<inline-css>true</inline-css>

	<!-- The timeout before showing the loading indicator.

	   The value is specified in ms.
          -->
	<indicator-timeout>500</indicator-timeout>

	<!-- The timeout for processing a double click event.

	   The value is specified in ms.
          -->
	<double-click-timeout>200</double-click-timeout>

	<!-- Ajax user agent list

           Wt considers three types of sessions:
	   - Ajax sessions: use Ajax and JavaScript
	   - plain HTML sessions: use plain old server GETs and POSTs
	   - bots: have clean internal paths and no persistent sessions

           By default, Wt does a browser detection to distinguish between
	   the first two: if a browser supports JavaScript (and has it
	   enabled), and has an Ajax DOM API, then Ajax sessions are chosen,
	   otherwise plain HTML sessions.

           Here, you may indicate which user agents should or should
           not receive an Ajax session regardless of what they report as
	   capabilities.

           Possible values for 'mode' or "white-list" or "black-list". A
	   white-list will only treat the listed agents as supporting Ajax,
	   all other agents will be served plain HTML sessions. A black-list
	   will always server plain HTML sessions to listed agents and
	   otherwise rely on browser capability detection.

           Each <user-agent> is a regular expression.
	  -->
	<user-agents type="ajax" mode="black-list">
	  <!-- <user-agent>.*Crappy browser.*</user-agent> -->
	</user-agents>

	<!-- Bot user agent list

           Here, you can specify user agents that should be should be
           treated as bots.

           Each <user-agent> is a regular expression.
	  -->
	<user-agents type="bot">
	  <user-agent>.*Googlebot.*</user-agent>
	  <user-agent>.*msnbot.*</user-agent>
	  <user-agent>.*Slurp.*</user-agent>
	  <user-agent>.*Crawler.*</user-agent>
	  <user-agent>.*Bot.*</user-agent>
	  <user-agent>.*ia_archiver.*</user-agent>
	  <user-agent>.*Twiceler.*</user-agent>
	</user-agents>

	<!-- Configures different rendering engines for certain browsers.

           Currently this is only used to select IE7 compatible rendering
	   engine for IE8, which solves problems of unreliable and slow
	   rendering performance for VML which Microsoft broke in IE8.

           Before 3.3.0, the default value was IE8=IE7, but since 3.3.0
	   this has been changed to an empty string (i.e. let IE8 use the
	   standard IE8 rendering engine) to take advantage of IE8's
	   improved CSS support.
	  -->
	<UA-Compatible>IE8=IE7</UA-Compatible>

	<!-- Configures whether the progressive bootstrap method is used.

	   The default bootstrap method first senses whether there is Ajax
	   support, and only then creates the application.

	   The progressive bootstrap method first renders a plain HTML
	   version and later upgrades to an Ajax version.

           (Not to be confused with the Twitter Bootstrap theme)
	  -->
	<progressive-bootstrap>false</progressive-bootstrap>

	<!-- Set session-ID cookie

	   In its default (and recommended) configuration, Wt does not
	   rely on cookies for session tracking.

	   Wt has several mechanisms in place to prevent session ID stealing:
	   - for an Ajax session, the session ID is not shown in the URL,
             avoiding session ID stealing through a referer attack.
	   - in case the session ID is present in the URL (e.g. for a plain
	     HTML session), Wt will redirect links to images or external
	     anchors through an intermediate page that censors the session ID

	   In case of the loss of a session ID, the impact is minimized:
	   - a full page refresh is not supported if the client IP address
	     changes or the user-agent changes
           - an Ajax update is protected by other state which is not exposed
	     in the URL

	   To still enable a full page refresh when the client IP address
	   changes, an additional cookie may be set which is used only
	   for this purpose, and can be enabled using this setting.
	  -->
	<session-id-cookie>false</session-id-cookie>

	<!-- Runtime Properties
     
           These properties may be used to adapt applications to their
	   deployment environment. Typical use is for paths to resources
	   that may or may not be shared between several applications.
	  -->
	<properties>
	    <!-- baseURL property

	       The absolute URL at which the application is deployed
	       (known to the user). This is needed only in two scenarios.

	       a) use of relative URLs in included XHTML

	       When you use relative URLs for images, etc... in
	       literal XHTML fragments (e.g. in WTemplate), which need
	       to resolve against the deployment path of the
	       application. This will not work as expected in the
	       presence of an internal application path.  This URL is
	       set as base URL in the HTML, against which relative
	       URLs are resolved so that these work correctly
	       regardless of the internal path. It is also used
	       explicitly in any URL generated by the library.

	       b) widgetset mode deployments

	       Another situation in which you need to define the baseURL is
	       when deploying a widgetset mode application behind a reverse
	       proxy. A widgetset mode application uses only absolute URLs
	       because it may be hosted in a web page from an entirely
	       different domain.

	       By default, no baseURL is specified, in which case Wt will
	       avoid using absolute URLs. Relative URLs passed in API calls
	       will be "fixed" so that they resolve against the location of the
	       application deploy path, even in the presence of an
	       internal path.
	      -->
	    <!-- <property name="baseURL">"http://mysite.com/app</property> -->

            <!-- resourcesURL property

	       The URL at which the resources/ folder is deployed that
	       comes distributed with Wt and contains auxiliary files
	       used to primarily for styles and themes.

	       The default value is 'resources/'
              -->
	    <property name="resourcesURL">resources/</property>

	    <!-- extBaseURL property

               Used in conjunction with Ext:: widgets, and points to the
	       URL of Ext JavaScript and resource files (css, images).
	       See the documentation for the Ext namespace for details.

	       The default value is 'ext/'
              -->
	    <property name="extBaseURL">ext/</property>

	    <!-- favicon property

	       By default, a browser will try to fetch a /favicon.ico icon
	       from the root of your web server which is used as an icon
	       for your application. You can specify an alternative location
	       by setting this property, or for an individual application
	       entry point by passing a location to WServer::addEntryPoint().
	      -->
	    <!-- <property name="favicon">images/favicon.ico</property> -->
	</properties>

    </application-settings>


    <!-- Override settings for specific applications.

       Location refers to physical filesystem location of the
       application. The application prints this location (which
       corresponds to argv[0]) to the log file on startup, and this
       should match exactly.
      -->
    <!--
    <application-settings
       location="/var/www/localhost/wt-examples/hello.wt">
    </application-settings>
    -->
</server>