Streamed live on Jun 14, 2017 Rchain Community Debrief 29
1:00 Greg Welcomes everyone https://youtu.be/mHnACqYhlig?t=60
2:32 Kent talks about Rosette, Rholang and Hello World https://youtu.be/mHnACqYhlig?t=152
3:33 Greg talks about the Rholang tutorial being produced and praises both Kent's ability to work with Rosette and Rholang and the project's inherant flexibility. https://youtu.be/mHnACqYhlig?t=213
8:19 Mike Stay talks about two papers on N-Category Cafe. https://youtu.be/mHnACqYhlig?t=500
19:50 Greg comments on Mike's discussion of the n-category cafe papers,fresh names and Galois extensions https://youtu.be/mHnACqYhlig?t=1180
23:49 Ed talks about floating point versus bignum. https://youtu.be/mHnACqYhlig?t=1429
24:58 Mike talks about standards https://youtu.be/mHnACqYhlig?t=1500
25:30 Greg and others talk about housekeeping-lagging https://youtu.be/mHnACqYhlig?t=1530
27:38 Greg talks about brain candy in Rholang Conway's Surreal Numbers https://youtu.be/mHnACqYhlig?t=1658
28:21 Ed talks about document that looks at supply of staking tokens for investors, board resolution of the cooperative. Investors are approaching Greg and Ed. Fundraiser pitches. Staffing forecasts. https://youtu.be/mHnACqYhlig?t=1701
31:05 Greg comments on staffing. https://youtu.be/mHnACqYhlig?t=1865
32:53 Ed talks about tax considerations, Identity, Holding co. take on identity, brave browser and civic IPOs, attention economy. Greg and Joseph add comments about attention economy. Nathan talks about animated video production and branding. https://youtu.be/mHnACqYhlig?t=1973
38:50 Greg talks about Lisa taking on the CFO role. Mentions they are prototyping the LADL algorithm. Investor meeting with Otto Capital. Talks about using Rchain as a metadata search tool. Mike joins this discussion. https://youtu.be/mHnACqYhlig?t=2330
48:50 HJ offers updates on activists/github. 40 actively involved. Mentions accidental exposure of personal information. Greg comments. Evan's advice will be sought. Ed joins discussion. Mike expresses concern. "Chief privacy officer." Board will compose a privacy policy. Navneet makes suggestions on permissions. Jim makes plug for noob-school. https://youtu.be/mHnACqYhlig?t=2930
54:25 Joseph D. gives update on pricing of Rchain services in Rholang. Dynamic mechanism vs. static. "an overlap in what our static versus dynamic cost analysis... they kind of require each other" Long discussion of pricing mechanism between Joseph, Greg, Navneet, and Mike. https://youtu.be/mHnACqYhlig?t=3265
77:50 Greg Announces he is getting married! Congratulations! https://youtu.be/mHnACqYhlig?t=4670
Raw transcript follows. Needs proofreading.
1:19 Christian I think Ed's going to join us 1:22 as well so I'll wait before we get 1:30 started 1:50 all right I don't know if he's okay well 1:57 let's just go ahead and get started and 2:00 and we have a lot a lot to talk about 2:02 but hopefully ed will join because we 2:05 also had some board resolutions that we 2:07 wanted to mention mm and let's go ahead 2:15 and the technical the technical updates 2:18 are quite fun Kent has made a lot of 2:22 progress with the compiler um so 2:29 about what you've been working on and we 2:31 can kind of go through that together 2:34 sure um so I guess first off we have the 2:42 we we cleaned up some of the tuple 2:45 space input some of the produce consume 2:49 functions on in the tuple space 2:52 implementation and then um and then we 2:57 went on and implemented to be um the I 3:06 guess the comparison and the arithmetic 3:09 operations 3:11 um on the compiler side and yeah that's 3:18 I think that's the sort of major updates 3:23 yeah yeah that's done that's exactly 3:26 right so so so look they're a couple of 3:30 little things that we want to mention 3:31 but this is all in preparation so at the 3:33 end of the month beginning on the 26th 3:35 Kent Joseph Alex Balkan and hopefully 3:38 Andrew Watson and I will all be working 3:41 intensively to get out a tutorial on 3:45 programming and rolling so there we're 3:49 going to get out kind of hello world 3:51 examples of wallet and a token contracts 3:59 so how you would write those and rolling 4:02 and of course for those kinds of things 4:05 you need arithmetic and you need unique 4:10 comparators you you need to be able to 4:13 say add this much to the existing 4:15 balance unless the balance is less than 4:18 zero or take this from the existing 4:22 balance less than zero so those are 4:25 those are basic kinds of things you want 4:27 to be able to say and so those the 4:30 syntax for those is is now available 4:33 inside the language and there are a few 4:36 examples that that Kent has put up I in 4:41 preparation for this 4:42 step I put up some work both in the 4:46 rolling spec but also in December 4:49 January time frame I wrote down how this 4:52 relates to our formal semantics so what 4:55 I what I showed was that you can do a 4:59 full faithful Church Church numeral 5:05 style in coatings and and I covered not 5:09 just sort of the arithmetic kinds of 5:11 operations but operations for for lists 5:16 and other kinds of collections so so 5:19 basically all of the the built-in data 5:21 types have a nice and friendly 5:25 interpretation just in the core language 5:28 which is why we could write down the 5:29 core language and say you know that's it 5:32 that's everything 5:32 um but of course we know that that from 5:36 the point of view of convenience of 5:37 programming it's it's not nice to do 5:40 Finance using Church numerals um so so 5:48 that that is why you have to... why the 5:53 language needs to support some higher 5:54 level expression languages and sort of 5:57 many expression languages inside the 5:58 language for those kinds of computations 6:02 and calculations now what makes this 6:04 really interesting just from a language 6:06 design perspective is that because 6:10 processes are first-class so that means 6:14 that processes can be passed around and 6:17 processes can be stored and these kinds 6:21 of things it makes it easier to then say 6:25 that values or processes so in 6:29 particular you can say things like the 6:35 number one or the string hello world 6:37 those are all those are all values and 6:41 because of the way a row line works 6:44 there they end up being processes and 6:48 and so we have a natural way to provide 6:53 the common 6:55 the the sort of common arithmetic syntax 6:59 which is a little bit harder in and 7:02 other kinds of approaches it the way I 7:07 like to think about it is because of the 7:08 reflective character 7:12 of Rholang we we have a relationship we have 7:17 sort of that the multiset ends up being 7:19 the data structure for role I like the 7:22 list is the data structure for Lisp or 7:26 scheme or those kinds of which is we 7:33 sort of have the same kind of natural 7:34 data program relationship inside Rholang 7:38 and the ease with which Kent has been 7:42 able to put that in is is both a 7:46 testament to Kent and his skills and and 7:49 a native intelligence but also to the 7:53 cleanliness of the semantics so so I 7:57 heartily encourage everyone to go and 7:59 take a look at that and along the lines 8:03 of semantics we also had a couple of 8:05 nice blog posts that Mike put up on the 8:10 n-Category cafe and Mike I was 8:12 wondering if you wanted to talk a little 8:13 bit about those since that relates to 8:16 the typing system for a length. sure so 8:19 we wrote up those two papers and put 8:21 them on the archive and that garnered 8:24 some interest from John Baez at the end 8:27 category for cafe so I wrote up one post 8:30 that introduced the concept of a graph 8:35 and with career theory for programmers 8:39 of the veer theory is very similar to an 8:43 interface for a class that is it gives 8:46 the API but not any of the 8:50 implementation details together with 8:53 axioms that any implementation has to 8:56 satisfy so then an implementation of 8:59 that would consist of code for the the 9:03 methods and so on 9:05 together with proofs that the code 9:07 satisfied those axioms in most 9:12 programming languages people will 9:16 try to substitute tests for proofs but 9:18 you know this is a very mathematical 9:20 thing so we stick with with proofs so 9:24 graph enriched Lawvere theory captures 9:29 captures the basic data that goes into 9:33 describing a term calculus so the first 9:37 thing that you need in order to describe 9:39 how a computer computes is what the 9:41 state of the computer looks like if 9:45 you're doing a simple computation model 9:49 like lambda calculus the entire state of 9:51 the computer is just the lambda term if 9:54 you're doing computation on a Turing 9:58 machine however you have the state of 10:00 the tape you have the transition table 10:07 that shows how writing a cell here I'm 10:10 sorry if you're it if you're pointing it 10:12 at this cell here then if it's a zero 10:15 you move to the left that was already 10:17 change it to a one and move to the right 10:18 or if it's one then you you know do 10:24 whatever right that the transition table 10:26 of the Turing machine also the current 10:33 state of the Turing machine and the 10:38 position of the read/write head on the 10:41 tape so all of that stuff has to go into 10:43 the description of the current state of 10:45 the Turing machine. Real programming languages 10:48 have immensely more complicated states 10:52 in JavaScript you've got two different 10:55 stacks a heap and a whole bunch of other 11:00 stuff that that goes in there into 11:04 describing the state so for each 11:06 different piece of this state you'll 11:08 have what's called a sort that 11:12 represents that piece the next chunk of 11:17 data you need is how to build up each 11:21 chunk so in lambda calculus you'll start 11:25 with variables so you'll have one sort 11:28 for variables 11:29 and then you build up terms by 11:31 application and abstraction so there are 11:34 these term constructors that you 11:36 describe next there are equations on the 11:43 term constructors so that tells what 11:47 kind of structure that is written down 11:50 in the syntax we want to ignore for 11:53 example if we're writing a function f of 11:57 X equals x squared that's really the 11:59 same function as f of y equals y squared 12:01 we don't care which variable is used 12:04 it's a dummy variable 12:05 similarly in rho calculus we don't care 12:08 about the order in which processes are 12:13 written down in the term because they're 12:16 all running in parallel there there's no 12:19 implicit order on these processes so 12:22 writing down equations that Express what 12:28 kind of structure we want to ignore is 12:30 the third piece the fourth piece is the 12:35 reduction rules the lambda calculus 12:38 that's beta reduction in rho calculus 12:41 that's the synchronization and 12:44 communication rule 12:48 and in and finally there is a machine 12:55 that's the transition yes that's the 12:57 transition function for a virtual 12:59 machine and finally there's the set of 13:03 contexts in which reductions can happen 13:06 in most modern functional programming 13:09 languages even if you have T reducing to 13:13 T Prime some term some state of the 13:15 computer T reducing to T Prime 13:18 you don't want lambda X T to reduce to 13:23 lambda X T Prime in Haskell literature 13:26 that's called week head normal form that 13:28 you don't reduce anything underneath a 13:32 lambda so all of that data is captured 13:38 in this notion of a graph enriched 13:41 lawvere theory so that was what the first 13:44 post is about the second post was about 13:46 how we can use reflection which is the R 13:51 in row calculus to make reasoning about 13:56 the operational semantics easier it 13:59 turns out that reasoning about languages 14:01 with binders in it is hard in 2005 there 14:05 is a set of challenge problems on 14:10 representing binders called the POPLmark 14:13 challenge but the original approach 14:17 to reasoning about languages with 14:19 binders was to eliminate them using some 14:23 kind of 14:25 homomorphism of calculate so in the case 14:29 of lambda calculus to get rid of the 14:32 lambda binder you translate it into 14:38 excuse me into SKI calculus where 14:40 they're just these three combinators and 14:43 you can express every lambda term in 14:47 terms of those and so that result had 14:55 been known since the thirties that you 14:58 could do this abstraction elimination 15:01 when Milner proposed the pi calculus in 15:04 1990 people started trying to do 15:07 something similar Yoshida in 1995 was 15:10 able to get rid of binders coming from 15:13 an input prefix so the before 15:16 construction in rolling but she wasn't 15:21 able to get rid of the new names sorry 15:25 that the bound names coming from the new 15:27 operator Greg and Matthias Radestock 15:33 matthias or matthias anyway 15:39 mattias MTS thank you Matthias Radestock 15:43 showed how to use quoting and 15:48 dereference operators to eliminate the 15:53 use of new in the PI calculus and in 15:56 their paper that described the rho 15:59 calculus from which rolling is derived 16:02 and so then in our papers we used that 16:08 technique in two different ways on pi 16:13 calculus 1 using the Yoshida style 16:16 abstraction and another one using s ki 16:19 style abstraction elimination 16:24 and so I described the resulting 16:26 theories there so that's a summary of 16:31 the lead-up work we're doing the the 16:35 resulting graph and rich love your 16:37 theories capture the operational 16:38 semantics and then you can use those to 16:42 in applicative calculate like lambda 16:46 calculus and SK I you can throw away 16:49 some information and recover the 16:53 denotational semantics that is that 16:58 individual terms can be modeled by 17:01 functions in concurrent calculate like 17:04 row calculus and pi calculus it the 17:08 evolution is of course non-deterministic 17:10 so you can't model them directly by 17:13 functions and this gives us a roadmap 17:18 for how to how to approach the 17:22 denotational semantics of these 17:24 concurrent calculate I can ask really 17:28 quick Mike 17:29 okay what's specific about the quoting 17:33 and dereference operators that makes 17:35 that possible it's the fact that new can 17:44 be thought of as splitting up into two 17:46 parts one of them is very much like an 17:52 input in that it's declaring a new scope 17:55 under which a name is bound and the 17:58 other part is providing a fresh name so 18:01 new can be represented as input a fresh 18:06 name from some process that produces 18:08 them and then evaluate the this term in 18:12 that context and that's precisely what 18:16 what Greg's transformation does is come 18:22 up and it's parametric in two processes 18:26 one of which defines a name space and 18:30 the other one produces a fresh name in 18:33 that name space 18:37 so as you recurse downward through the 18:42 structure of the term it's defining a 18:45 namespace for each substructure of the 18:47 term and then defining a new fresh name 18:51 in that namespace so that it's 18:53 guaranteed to be fresh in netskope 18:57 actually first of all I want to say that 19:00 was a remarkably succinctly clear 19:02 description thanks like that was this 19:04 really you did it better than I did um 19:10 but I just wanted to panel ambient 19:13 calculus does that lot have input 19:15 prefixes there's there's an extension to 19:18 the ambient calculus that does but the 19:21 the pure and being calculus without that 19:26 doesn't have prefixes and so that makes 19:28 it very hard to use reflection to do 19:32 this trick Greg had to come up with an 19:35 entirely different approach to ambient 19:39 to get a representation of them in rho 19:43 calculus using namespaces for the 19:45 ambience rather than just the standard 19:48 reflection that he used to get rho 19:51 calculus from pi yeah and actually is 19:56 quite interesting there 20:01 will distract encoding of the ambient 20:02 calculus into PI so showing that the ambient 20:06 calculus is considerably more expressive 20:10 than pi calculus the ambient calculus is 20:14 like at the other end of the spectrum 20:16 it's too expressive one one thing that 20:22 is what one thing that is is worth 20:25 mentioning also just as a quick 20:27 extension to to Mike's comments when 20:31 Matias and I were doing this work Matias 20:34 was sort of exploring an obvious 20:37 solution which is that you like what one 20:41 suite once we knew we had this trick 20:43 then then there there essentially to 20:45 this bifurcation in the approaches one 20:48 is you have a fresh name server so this 20:52 is kind of a runtime approach when you 20:54 need a fresh name 20:57 the fresh name server will 20:59 pony up a fresh name for you and even in 21:03 2002 when I was carrying out this work I 21:07 didn't like that approach because it's 21:11 centralized and I was already aiming for 21:15 a decentralized solutions so the 21:18 solution that I came up with in the 21:26 um paper is a compile-time solution with 21:30 which is decentralized so there is no 21:35 ponying up these fresh names it 21:38 happens as a result of a compilation of 21:41 the of the PI term so 21:45 erm you compile it into a rho calculus 21:47 term and as a part of the process of the 21:49 wrote of the rho calculus the 21:51 compilation two rho calculus you you get 21:55 this decentralized fresh name idea and 21:58 that that turns out to be quite powerful 22:01 and it's almost like the future was was 22:07 reaching back into my mind and second 22:10 saying you know you're gonna need this 22:12 decentralized stuff soon so so that's 22:17 that that is one of the one one of the 22:21 subtleties there there's a there's 22:22 another subtlety which is super super 22:24 mathematical so I will go into it but 22:26 but for those out there who are 22:28 interested there's a connection between 22:29 fresh names and Galois extensions so 22:32 there's a third there's a third solution 22:34 and the other way to think about is that 22:36 it's related to being able to have 22:40 recursively defined names so the grammar 22:44 for the rho calculus does not allow for 22:46 there to be infinitary names all the 22:50 names have the finite information in 22:52 them but if you but you can provide 22:57 recursive limit points and and then 23:01 there's a there's a perfectly consistent 23:02 calculus that's representable in a 23:04 computer that has recurred 23:11 and that gives you yet a third approach 23:13 to to eliminating new I haven't written 23:17 up that approach so if there's any 23:19 intern summer interns out there who want 23:23 to who want to work on that that 23:25 particular extension where we're always 23:27 we're always up for folks who are 23:29 interested in that kind of stuff 23:30 engaging with folks that way um thank 23:32 you Mike that was fantastic 23:34 um so is it on I don't know yeah I'm 23:39 here 23:39 alright well let's let's move on to - 23:45 from the technical stuff - a little bit 23:47 of the business stuff and then come back 23:48 to some other technical updates oh yeah 23:50 I got a few things to update on the 23:54 first thing is a actually a question for 23:58 Kent or Greg on the the math in in 24:03 rolling uh it was I'll tell a little 24:06 story is like so Lisa our CFO was 24:10 working on balancing out you know some 24:14 of the historic amp and Bitcoin 24:16 transactions for coop and noticed that 24:21 Excel didn't add up correctly when it 24:24 went down to eight digits of precision 24:27 and actually what happened is Excel and 24:31 Google sheet used floating point math 24:33 which only handles fifteen significant 24:36 digits so it was of course that you know 24:42 for CFO types that like to balance to 24:45 the penny this was driving her nuts and 24:47 I've been so you know for me I ever 24:50 realized the problem I thought it would 24:52 just round it you know she's like no I 24:53 can't do that 24:55 so my question 24:59 but while IBM was involved in the ECMO 25:03 script standardization yeah committee 25:06 for es4 and the only thing they wanted 25:09 the one and only feature and all of 25:13 their support was focused on give us 25:15 decimal number types yeah so yeah I 25:18 think you know either fixed point 25:20 precision or anyway we just we have to 25:22 make sure 25:23 we handle many significant digits I'm 25:29 some bloody fun so we're doing this in 25:34 stages we're just just just too quick 25:41 your audios cutting over can you hear me 25:43 your audios been cutting out all call no 25:46 one else's has well that's weird 25:49 I wonder what's going on it's been 25:51 having actually let me see what's going 25:56 on let me try sir Greg it's like this 26:00 every time in when we're in one of these 26:02 zoom meetings 26:03 so it's unlikely that you'll be able to 26:05 fix it right yeah it's a it happens in 26:07 Skype occasionally do I believe any with 26:11 your machine Greg the simple solution 26:13 table yeah I'm not sure actually kit can 26:16 you am i cutting out now can you hear me 26:18 okay now it's something out you know 26:22 very frequently a Greg mean every time 26:24 you talk euro is use on for two seconds 26:27 and then it goes out for two seconds 26:29 okay well is it better now can you hear 26:32 me now at the most okay cool so what I 26:35 just did was I switched from my to the 26:39 five on the network provided by Comcast 26:43 so I think it may be a network issue 26:45 with the to the my this machine is just 26:48 far enough away from the modem that I I 26:50 usually have it on the - rather than the 26:52 five ah so so but I just switched to the 26:55 five and that seems to be better 26:57 um can you is it is has it stayed goes 27:00 to the bar all right Austin so it seems 27:02 like it was a networking issue um so I 27:07 just wanted to mention briefly what 27:09 we're doing obviously we're doing the 27:11 language in stages right we're you know 27:13 we start with a core that we will well 27:15 understand then we then we make 27:17 extensions so the extensions for the 27:21 numeric types right now are just the 27:23 numeric types that are easily supported 27:25 in Java and so so we're going to be 27:30 stuck with those limitations however one 27:33 of the little pieces of brain candy that 27:36 I put in the in the the church numeral 27:42 paper for rolling is that we can not 27:47 only do infinite precision we can do we 27:50 can do one better 27:52 we have we have um there's there's a 27:56 clean and coding of Conway's surreal 27:59 numbers into rolling no I'm not claiming 28:03 that that would be computationally 28:04 efficient but we can be infinitesimally 28:09 precise Oh 28:12 so anyway back back to you that's 28:16 awesome that's good to hear um so I'm 28:19 gonna jump around on different topics uh 28:21 as as Greg reviewed last week we worked 28:26 on a document around the supply of the 28:30 staking tokens and I'll face the link 28:34 here ins is Spock if I find the right 28:36 place where is it hiding there is or in 28:40 zoom rather so that that's the document 28:43 and one of the reasons that's important 28:46 is for new investors that we're talking 28:49 to want to really understand you know 28:53 that either either the amount of rock 28:58 they may get or certainly the amount of 29:00 rocks that are chain holdings holds and 29:04 what its future utility is relative or 29:08 they know the conversion to Reb what 29:09 that future utility is relative to other 29:12 staking tokens that might be produced 29:15 and we talked about that last week and 29:16 so feel but anyway so we're going to 29:18 have a board resolution from the 29:20 cooperative so that they have you know 29:23 as much comfort as as we can provide in 29:26 that in their decision um let's see so 29:31 yeah a number of investors have been 29:33 approaching a Greg and me a so often 29:37 separately sometimes together and we're 29:39 actively following up with those 29:41 conversations um we are continuing to 29:46 have fundraiser pitches and feedback we 29:48 have another session this afternoon and 29:51 those are going quite well as you all 29:54 know it takes a while to explain um 29:56 depending on you know who the investor 29:59 is and what their comfort level is with 30:01 crypto and so forth so that we have to 30:04 be a little bit flew 30:05 in how we present you know based on 30:08 their background but also just where the 30:10 conversation goes because the surface 30:14 area to cover is so large we had a Glade 30:19 a good financial planning session Lisa 30:22 Greg and navneet and myself earlier this 30:24 week on Monday I think that was where we 30:29 talked about the the staffing forecast 30:32 of the coop growing the team to about 20 30:38 for the development and then a few more 30:41 those would be as as many contractors as 30:45 possible versus employees is kind of the 30:49 the mix and that would allow if I 30:51 remember Greg about four or five on each 30:55 of the teams for storage for comms for 30:59 the virtual machine layer and for system 31:03 contracts I may not have that quite 31:05 perfect but that's roughly the shape of 31:06 it yeah that's from that's roughly the 31:09 shape of it and then things will morph 31:11 over time right like like the comms and 31:13 storage are fairly well contained 31:16 efforts and so you know once those are 31:20 nailed down they're nailed down and so 31:22 people if people want to be moving 31:24 upward to working on system level 31:26 contracts or they want to try their hand 31:27 at some of the finer details of language 31:29 design or compiler worker or they want 31:32 to get into the consensus stuff the 31:34 Casper stuff you know um you know there 31:36 is more than enough work and and and 31:41 it's it's rich interesting and and good 31:44 technical work so let's go ahead and 31:46 then put that out there so right now 31:49 Nash and Kelly and Mike are with their 31:53 with power of X and their power of X is 31:55 handling our staffing so if there are 31:59 developers out there we kind of have to 32:02 hit the ground running as soon as the 32:05 the major bulk of the first tranche hits 32:07 the bank accounts we need to be rolling 32:10 so we need to we need developers you 32:13 know in interviewing best right right 32:16 away so if you're out there you're just 32:18 encrypt occur 32:19 see you're interested in language design 32:20 you're interested in consensus these are 32:23 all tough interesting fascinating 32:25 problems and a either desire and 32:29 aptitude to learn formal method just and 32:31 functional programming is part of that 32:33 whole mix yeah concurrency theory 32:35 absolutely yeah awesome um so can 32:42 continue please 32:44 all right cool uh so obviously on the 32:47 financial planning side that there's you 32:49 know more to do um the the 32:52 considerations around tax are really 32:54 interesting uh housing planning like 32:58 interesting interesting and maddening at 33:01 the same time know how the how the you 33:06 know that the assets are deployed in the 33:09 various crypto currencies and US dollars 33:12 so that's a that's an interesting topic 33:14 as well um so that's moving ahead uh and 33:18 then I wanted to mention also I've 33:21 mentioned this every week but we're 33:22 continuing our exploration on identity 33:25 related things so Glenn and I Quinn's on 33:28 the call here of participated in the 33:32 co-op's of bring your own identity 33:34 session that Jim White's Garber's has 33:36 been leading a slab of Saturday morning 33:38 and at least I think Glenn will be 33:41 continuing to attend those I'll tend 33:42 occasionally so that those are great 33:44 conversations um you know on the holding 33:47 side where we're looking at this 33:48 intersection of you know identity of 33:52 wallets and browser kind of interactions 33:55 and so for example I looked at you port 33:59 this week and kind of understood how how 34:02 that was working um other things that 34:04 happened within well I guess it was 34:06 maybe two weeks ago already 34:08 you know the brave browser you know 34:10 raised thirty five million in ICO in ten 34:12 minutes 34:13 um Civic just two days ago or maybe it 34:16 was yesterday thirty three million 34:18 dollars most of it was sold even before 34:20 the ICO happened um so is essentially 34:23 instantly committed you know so these 34:27 are all kinds of similar place where a 34:29 brave and civic are 34:31 working that the user can own and manage 34:34 their identity and growing it I assume 34:37 their strategies to grow into a more 34:39 rich self sovereign identity but also 34:42 you know for micro payments or to you 34:48 know for for those to reduce the 34:50 barriers and frictions between a Content 34:54 producer and the consumer so that 34:56 there's you know removing all the 34:59 middlemen and leeches along the way so 35:01 that those that are consuming content 35:05 can can get that pay for that content 35:10 and a benefit directly to the producer 35:14 so that's kind of common theme of these 35:16 things as well as the attention economy 35:18 features that we ultimately want to 35:20 build so I am really quick just a slight 35:26 commentary on the attention economy and 35:29 Greg a question for you when did you 35:32 first publish the paper on the attention 35:34 economy so the concept of the attention 35:36 economy actually uh you know wasn't 35:39 invented by scenario or Greg others you 35:43 know had that that term and that concept 35:46 out there not necessarily related to 35:48 crypto the nitrogen attention economy it 35:53 goes because way well back 35:56 I remember reading popular press 35:59 articles about attention economy in 2007 36:05 and 2008 some people were a lot of 36:09 people noticed that the the social media 36:12 constituted sort of first mechanization 36:16 or realization of the idea of attention 36:19 economy or some of the some of the key 36:21 characteristics of attention economy so 36:23 yeah that notion goes way back now 36:25 though the the specific details of you 36:31 know how this relates to cryptocurrency 36:33 how this relates to micro transactions 36:35 and and these kinds of things those were 36:39 obviously much later much later in the 36:42 game yeah 36:44 the well my point in asking was was more 36:47 so unlike the the peer-to-peer content 36:49 delivery and marketing side was you know 36:53 especially in the past couple years 36:54 there's been an absolute explosion of of 36:59 entities that are leveraging the 37:01 peer-to-peer marketing idea to further 37:04 their own brands and I thought it was 37:06 really really cool that now the industry 37:09 so well corresponds with what the 37:12 attention economy was essentially built 37:13 for low estate exactly once one micro 37:16 payments become more and more practical 37:18 on the blockchain with with our team 37:20 then it the world's gonna change all 37:23 right please yeah silligan going back to 37:26 my status let me finish up the status 37:28 and then we give Ana back to Greg so the 37:31 the other thing is so we have a medium 37:34 article and that's been written of 37:37 introducing our chain Shinkai Thornton 37:39 is written and he's going to get that 37:41 posted via KOIN fund site and then in 37:44 the near future which is great it's 37:46 going to get a lot of eyeballs on it and 37:49 then also Nathan are either do you want 37:52 to give an update on the animated video 37:54 yeah 37:56 the animated video is and it's it it's 37:59 almost cooked or early complete we have 38:01 the full video we're just going to edit 38:02 a bit of the animation itself and and 38:07 then also get feedback internally about 38:10 it so that we can go forward in making 38:14 other ones to address different 38:18 communities both developer and non 38:20 developer it and then investors and then 38:23 lastly about in that zone is we're also 38:27 in the middle of figuring out the 38:28 branding document and the colors that 38:31 we're going to use the token logo and 38:37 basically all things sort of branding 38:40 which we did not have going into the 38:43 video so that's the update awesome 38:45 thanks Nathan 38:46 uh back to you Greg uh thanks ed you're 38:53 welcome we need we need to I need to 38:57 have a big you know 38:58 world map behind me aren't you at 5:00 39:01 that's correct 39:04 exactly um so yeah some a few other 39:10 updates one of the part of the meeting 39:12 with Lisa was also to go over her 39:15 proposal to take on an acting CFO role 39:18 but the coop and so we've kind of shaken 39:23 hands on that um and so that'll be good 39:27 because I've got a lot going a lot of my 39:31 plate when it comes to just managing the 39:33 the business part I'd really like to 39:35 shuffle that off trying to get even just 39:39 getting the house in order so I can hand 39:41 over a big chunk of the functions to 39:44 Lisa nhj has been it's been a very very 39:48 time-consuming 39:50 just the big run up on on Bitcoin and 39:54 other cryptocurrencies has made it 39:55 really really hard to do certain 39:58 functions with respect to amps and the 40:00 omni platform so that was that's been 40:03 incredibly time-consuming so I'm 40:07 grateful to get some help on that um 40:10 let's see I also wanted to mention um 40:14 that we're now Mike and I sort of as a 40:19 culmination of some of the work that 40:21 Mike and I've been doing we're now 40:22 moving over into the prototyping stage 40:26 for the label algorithm so we've gotten 40:30 it to the point that we're satisfied 40:31 that the the mathematics hangs together 40:36 and so we're now making decisions about 40:41 the platform for the prototyping of the 40:45 the typing the typing algorithm and look 40:51 forward to seeing at those commits to 40:54 the repo in the not-too-distant future 40:57 I'm super excited about that I'm really 41:00 really excited this is uh you know I 41:06 don't know how to say this is gonna this 41:07 is gonna this is well let me put it this 41:10 way um 41:12 when we were we had a really nice 41:14 investor meeting with some of the folks 41:18 from auto capital on Friday kind of an 41:21 in-depth presentation and mostly we were 41:24 just focusing on our chain and the 41:26 crypto applications and these kinds of 41:29 things but it came to light you know so 41:31 we turned the conversation turned 41:33 towards the formal methods and we 41:40 mentioned the the ladle algorithm and 41:43 then I just you know offhandedly pointed 41:46 out that once you have a rich typing 41:49 mechanism like this it becomes a query 41:52 mechanism and their their eyes got wide 41:56 and the way the way to understand this 41:59 is think about the language I use 42:04 sometimes is the code is kind of the 42:05 dark matter of the internet what do I 42:09 mean by that well what I mean is that 42:12 it's a huge data asset right if you 42:15 think about github github has terabytes 42:19 of data maybe petabytes of data up in 42:22 there and and a big chunk of data Scone 42:26 and yet it's impossible to search that 42:29 code on the basis of what it does but 42:33 you have to you have to search that code 42:34 either because someone else knows that 42:36 code base so you search it socially or 42:39 you search it on the basis of some kind 42:42 of metadata or annotations which are 42:44 also often completely either misleading 42:47 or unsatisfying but if you couldn't 42:49 search it on the basis of what it does 42:51 um that would that would make it more 42:54 like a data asset now if you think about 42:57 a github is just the code that's public 43:00 plus some code that's private but it's 43:02 mostly you know public assets behind the 43:06 firewall there's a bunch of code in the 43:08 enterprises and that's also a data asset 43:11 that's completely unsearchable right all 43:14 of that all of that is searched on the 43:16 basis of social it's searched on you 43:18 know somebody who knows somebody who 43:19 knows somebody if you could search that 43:22 code just on the basis of what it does 43:25 um then certainly that code as a data 43:27 asset would be would be many times more 43:29 valuable the ladle algorithm gives us a 43:33 mechanism to search code on the basis of 43:36 what it does now if you if you move if 43:41 you think about the revolution that's 43:43 taking place with respect to contracts 43:48 and smart contracts this is not only 43:51 recovering the value of things from the 43:54 past which have been undervalued and 43:56 underutilized but it's also pointing us 44:00 towards the future because if our chain 44:03 succeeds or even you know if a theory 44:07 amuck see him succeed succeeds seeds 44:10 duration of contracts just like there's 44:14 been a proliferation of JavaScript 44:16 programs and Java programs and c-sharp 44:18 programs and and so on and so forth so 44:21 they're going to be a proliferation of 44:22 solidity programs and there's going to 44:25 be a proliferation of rolling programs 44:27 and now what you want to do especially 44:30 when we're talking about financial 44:32 models is you want to be able to manage 44:34 those models not just the data that they 44:36 work on but manage the models themselves 44:38 as if they were data assets and so you 44:43 want to be able to say remember that 44:44 contract that we use to to model 44:47 derivatives in this way right I can 44:50 trigger the name of it I really remember 44:52 who did it 44:52 right that's that's that's the sort of 44:54 scenario we can picture in our minds but 44:58 if you could write down a little type 45:02 base description of of that models you 45:06 sketch out roughly what it did then you 45:09 can actually ask the system to go and 45:11 search for that contract and so when we 45:13 when we describe some of these 45:15 potentials of just the ladle algorithm 45:18 alone um the investors got quite quite 45:22 interested in quite excited because they 45:24 hadn't they hadn't seen that angle that 45:26 the sort of you know yet another 45:30 dimension to to what we're doing with 45:33 with with our chain and rolling in the 45:36 ladle opera so 45:39 so that's why I'm very very excited to 45:42 be reaching this point where after after 45:45 almost a decade of research just on the 45:48 type the type part of this 45:50 we're getting to the point where we're 45:53 prototyping and that will be folded into 45:56 into the rolling a repo in the in the 46:01 upcoming months so it's very very 46:03 exciting for me personally but also I 46:05 think in terms of the the overall 46:07 offering and I really you know you 46:10 should you should allow your 46:12 imaginations to to run for a bit on this 46:15 they think about discovery protocols 46:18 based upon typing information so piecing 46:22 together piecing together contracts just 46:24 in time right that's another as another 46:27 example of the kind of thing that you 46:29 can do as a result of having the ladle 46:31 algorithm in operation I think Mike and 46:34 Nash have thought about a souped up 46:38 cloud-based linker how did you describe 46:44 it Mike well the idea was if you can 46:48 prove properties of a code base so you 46:54 know some subset of the programs on 46:57 github then when you do your query to 47:01 github you can get all of the code that 47:05 satisfies the property you're looking 47:10 what it means is that you can get three 47:14 monetization model or open source 47:17 s oftware where an open source developer 47:23 opts in to allowing his code to be 47:26 linked in response to one of these 47:27 queries and so the the linker service 47:33 gets a cut the open source author gets a 47:37 cut and the guy who links it in pays for 47:42 the privilege yeah that's that's a 47:45 beautiful beautiful model and and it's 47:47 tin abled not just by the the micro 47:50 currency capabilities at the blockchain 47:52 but it's critically enabled by the ladle 47:54 algorithm right it's only if you can 47:56 search on the basis of what the code 47:59 does that you have that this starts to 48:01 make more sense so so so that's that's 48:07 why we're super excited about that let's 48:09 see um any other updates that I can 48:13 think of I'm think I'm hey Greg I have I 48:19 have some real quick I need to bring to 48:20 your attention I needed I need to take 48:22 it off on between you and head though I 48:24 can do it in slack just side I just 48:26 needed yeah 48:28 that's pretty important it can be pretty 48:29 major maybe not so it's but it want to 48:33 take it here anyways keep going oh yeah 48:35 that's cool that's cool um yeah sure if 48:40 do you have time right after the call 48:41 yeah a little bit okay cool let's let's 48:44 do that um trying to think there's any 48:47 anything else I mean for me that's a lot 48:50 um have a few updates quick yeah go for 48:57 it oh yeah so then as you know I'm 49:04 searching for activities to do some side 49:07 work on the archer in cooperation and 49:10 there's no 5858 people registered and 49:17 forty people are actively involved on 49:20 our github our team members repository 49:23 there we collaborate together but 49:27 there's also some bad news 49:30 these people got registered for through 49:34 Google Forks and the answers of the 49:38 Google Forms are in spreadsheets and one 49:40 of those places was publicly exposed so 49:45 with some sensitive data that means more 49:48 than email addresses so I'm going to ask 49:52 Evan what to do about this because this 49:54 is a leakage of personal information 50:00 yeah I saw that yeah so so I'm glad with 50:05 brought to our attention and 50:06 we and that we're taking steps to 50:08 rectify the situation I understand that 50:12 the page has has been rigged given 50:15 different permissions so that that's a 50:17 step we can probably also figure out 50:21 exactly a sort of the level of traffic 50:25 t here and see if it was you know um you 50:30 know how far and wide that might have 50:32 spread yeah it was okay to the are chain 50:36 slacked on April 13th bringing this to 50:40 the attention of the community so it's 50:44 been publicly available at least um 50:48 April May and June I don't think that's 50:51 the case I pretty sure I checked 50:54 permission after that April comment and 50:57 and it was tightened for a while at 50:59 least yeah this through because it the 51:03 permission splint on and off so this 51:06 because of this Google Forms and well 51:10 giving access to it and it goes pretty 51:12 easy and it's a kind of failure in fact 51:15 but every old documents are now checked 51:21 and only certain people have access the 51:26 fact that it happened twice is kind of 51:27 concerning seems like we need to have 51:32 somebody involved in checking and 51:38 maintaining these things setting up 51:40 policies setting up procedures so that 51:43 that can't happen again things like that 51:46 yeah ah chief privacy officer or 51:49 something like that yeah yeah yeah it's 51:52 a good idea it's a good idea agreed 51:56 agreed 52:01 okay else H Jake 52:03 yeah the here the the co-op bylaws also 52:06 require that the the board pass a 52:09 privacy policy which we haven't done yet 52:12 I'd like to welcome Joseph to the to the 52:16 group I thought for the yesterday's very 52:18 excited 52:20 who wants to do something useful as a 52:22 lot to learn maybe in the noobs group 52:25 are only in that people wanting to run 52:30 rosette and run a scalar IDE and need a 52:37 little help and well who has somebody 52:40 step up steps up to coach that's prep 52:42 what we'll be doing tomorrow at one 52:44 o'clock New York time 52:50 Clemmie asking what we can do we can 52:53 pair up checking you know until we have 52:55 this a new person or you know people of 52:59 person I think can we do put two more 53:02 people as a verification anything we are 53:04 doing at least they can look into the 53:06 data and check the permissions 53:09 you 53:13 um permission permission sir proposing 53:18 to SJ is at least as they asked two more 53:21 people to look you know I mean we 53:24 shouldn't be any way doing that exposing 53:28 excel kind of thing but at least look 53:29 you know to the permissions by two more 53:32 people not only you and you know asked 53:34 to verify two more people so that the 53:37 permissions are correct they are not 53:38 public open and are you are you talking 53:43 about the UM Aramis no no no I mean 53:46 general document permissions where there 53:49 is a sensitive and function right and 53:55 yeah I I always make sure when I go in 54:00 and document that sensitive information 54:02 that the permissions are right I did 54:04 didn't notice the issues that he had um 54:16 I can I can give a an update on what 54:20 I've been working on these last few 54:22 minutes um cool cool all right so the 54:31 last time we spoke I think public 54:33 publicly about this the the route that 54:37 that we were pursuing was attaching 54:41 weights essentially to each of the teach 54:46 of the grammar constructs for rolling 54:50 and what we can always new and what we 54:54 definitely found out or you know and 54:57 what we definitely been seeing is that 54:59 on top of the e on top of the typical 55:05 information such as input size and such 55:07 as you know method invocation that we 55:10 cannot know statically at compile time 55:15 we're also dealing with non determinism 55:17 that's present just in the semantics 55:19 semantics of the language so the my 55:24 thinking 55:25 this past week has been kind of a step 55:27 beyond that and essentially essentially 55:31 you know the idea is that if we have a 55:34 dynamic mechanism then we can always get 55:36 the actual cost of a program if it's 55:40 dynamic if we have a static mechanism 55:43 then the best we can do is the maximum 55:47 and upper bound on the cost of 55:48 computation and you know as a small 55:52 example that's done typically there are 55:53 two types of non determinism one when 55:55 you have two output turns sending to one 55:58 input term and you don't know which one 56:00 which message is going to be received 56:02 first so in that case we will be forced 56:06 to take an upper bound on the largest 56:11 input size and the length of the process 56:14 of the continuation with that input and 56:18 then the other with the other message so 56:21 we take one out of those two paths and 56:24 then the other case of non determinism 56:25 of course when you have you know message 56:27 being said on one channel and there are 56:28 two input terms waiting to receive and 56:30 you don't know which is going to receive 56:33 and again we take max bounds on on that 56:36 term but since we can always get the 56:41 since we can always get a cost 56:42 dynamically that method should take 56:45 precedence developing developing that 56:48 method and actually it's pretty 56:51 interesting I'll explain in a moment but 56:52 we find an overlap in what our static 56:55 versus dynamic cost analysis they kind 56:58 of require each other so given that 57:03 given that when we when we run a program 57:05 dynamically we're going to have a set 57:07 amount of Rev and every action that we 57:09 take through the course of that program 57:11 is going to decrement the amount of read 57:13 that we have by a certain amount which 57:16 that action is assigned by the weight 57:18 and then the question is if I have a set 57:22 amount of Rev I have three processes 57:25 that are executing in parallel how can I 57:27 guarantee that all operations against 57:32 Rev are valid well there there are two 57:35 courses at least initially that 57:38 identified 57:39 one is to make rather a global value 57:42 which is a bad idea for a number of 57:45 reasons for performance and for for sake 57:47 of accuracy and that the second is is 57:51 actually to to make the cost analysis 57:55 statically by Max Val and then associate 58:00 that max possible cost with each process 58:02 and then use that as the amount that 58:05 we're deducting from dynamically with 58:08 the difference being the refund that we 58:10 get so dynamically we can think of P or 58:15 any any process as as a sequence of of 58:19 actions represented in bytecode and each 58:23 of those are going to affect some state 58:25 of the machine in some way most of those 58:28 will be internal unobservable and 58:32 they're just going to you know whatever 58:33 whatever you can think of a bytecode 58:34 doing and essentially we want to say 58:40 that if any point you reach a revenue 58:43 that is less than or equal to zero we 58:48 want a halt machine insufficient REV 58:50 but of course we also want our you know 58:54 our cost analysis to be consistent with 58:56 the operational semantics structural 58:58 congruences and by similarity of the 59:01 formal model so from from perspective of 59:05 bisimilarity we want to say that if 59:07 two processes can execute on top of of 59:12 course our you know our normal 59:14 requirements for four bisimilarity if 59:16 they can execute with the same amount of 59:18 Rev and both reach a stable halting 59:20 state which implies obviously that Rev 59:24 is not equal to not less than zero then 59:28 they are weakly bisimilar and we 59:31 obviously meet we can't say that they 59:33 are strongly bisimilar because we don't 59:39 know exactly because because they could 59:41 take different unobservable actions that 59:43 ultimately cause the same that have the 59:46 same effect on the environment that's 59:48 unobservable to whatever witness or to 59:50 the environment well 59:53 o h so you folded into your idea that 59:57 they that they do this the IO matching 60:00 right so when you set when you were 60:01 saying on top of the the additional 60:04 requirements of by simulation you so 60:06 that's where you're tacking in that they 60:08 do the IO matching right of course yeah 60:10 all right sorry okay yeah took 60:13 me a minute to parse what you were 60:15 saying there got it sorry about that 60:17 no well then well then the next question 60:20 then was you know how do we synchronize 60:23 essentially two processes how do we 60:25 actually you know transition the system 60:27 and generally the general idea is that 60:31 you know considering we have P part Q 60:33 output part input goes to Q prime or Q 60:37 with bound input that the system will 60:41 transition with the Rev left after 60:45 performing the input action and all of 60:49 course the remaining Rev in P and 60:52 remaining rev up to Q and when it 60:54 transitions it's going to copy the 60:57 remaining amount of Rev into the amount 60:59 of read that use Q witnesses and then 61:02 cue the continuation execute with that 61:04 amount I understand so you're looking at 61:09 the the problem of how to decrement the 61:11 Rev as it's executing yes okay okay yes 61:16 I wasn't quite I was able to follow 61:20 exactly so so yes that's it that's an 61:23 additional piece of the puzzle so in 61:26 addition to estimating the cost up front 61:31 which is a map from the contract to a 61:36 decorated contract so we also want to 61:40 look at the other side on when a 61:43 contract is executing and has a certain 61:46 amount of Rev what's the throttle right 61:50 like how far can it go but without 61:52 without reaching without really exactly 61:55 or a wheel I understand that you will 61:59 have the lower ceiling and then lower 62:02 bound and upper bound 62:03 I'm not sure what you guys are you know 62:06 talking in that sense right 62:08 no no what would Joseph is saying is you 62:12 you need to be able to provide a I'm 62:16 paraphrasing and so and so if I've 62:18 gotten it wrong please just jump in so 62:20 I'm just gonna echo back what I'm 62:21 hearing so eat one and I also want to be 62:27 a little bit more precise we're not 62:28 talking about Rev we're really talking 62:30 about phlogiston sorry legit sorry 62:33 Michael no no no it's totally 62:35 understandable um so a a party an agent 62:41 that wishes to execute a contract on the 62:45 Rho V M um supplies a certain amount of 62:51 Rev in order to get phlogiston the 62:54 question is how much Rev should they 62:57 provide and then once Li and the brendon 63:02 next the next question is once they 63:04 provided that Rev as the contract 63:07 executes and uses up logistic how is the 63:14 pledges in use data and how does that 63:17 relate to the decrement of the rev that 63:20 was supplied sorry um well anytime I 63:24 said rev please pretend I was saying 63:26 phlogiston what I meant was any like 63:28 essentially throughout the execution of 63:29 a process P any action that's internal 63:33 to that is going to decrement the amount 63:36 of phlogiston that we got because you're 63:37 affecting this state of the machine 63:39 essentially and in some way you know 63:41 whether that's whether you're increasing 63:43 the amount of memory that you're using 63:46 you know whether you're executing say an 63:49 instruction which has a cost 63:51 so either data or compute yeah so we so 63:54 we want to get kind of an idea of of how 63:57 much we're going to consume throughout 63:59 the lifetime of the contract so 64:02 statically at compile time we can take 64:04 the maximum like whatever the worst case 64:06 scenario is that takes the most amount 64:08 of phlogiston throughout the entire 64:10 execution and then we're going to 64:11 require that they pay that amount before 64:14 it's executed at amount that max 64:17 is then going to be used as our metric 64:19 dynamically that's going to be used as 64:21 the amount that we start with 64:22 dynamically so in that way we always 64:24 have enough logistic you know in the 64:27 worst possible scenario to finish but if 64:30 they overpay if you if you know say we 64:33 don't run at worst-case scenario then 64:35 we're gonna get a refund on the amount 64:36 of phlogiston that you gave us in red do 64:39 I want to step back you know I'm not 64:41 sure why you are including a term 64:43 dynamic right so if you are already 64:45 setting up a upper bound and lower bound 64:47 so let me step back further so somebody 64:50 at has certain amount of Reb with your 64:53 contract to be exited in the future date 64:55 whatever Millie and between Rev and Flo 64:59 distance right there will be some 65:00 mechanics how they can get certain 65:02 amount of lessons that will be 65:04 applicable to get the computation 65:06 storage and network all three things 65:08 together right sure yeah I'm not sure 65:11 how you dynamically changing because 65:13 before you put your contact into the 65:15 network my I would imagine you'll have a 65:18 fixed rate for these things which may 65:20 change with time but when you are 65:22 signing a contract that amount should 65:24 remain same you know so what what he's 65:29 saying is that as the contract executes 65:32 so let's consider a very simple contract 65:34 right so like the hello world contract 65:36 in the hello world contract the contract 65:40 waits on the channel world one for a 65:44 message now in order to run that 65:47 contract the agent that submitted the 65:50 contract for execution needs to supply 65:54 some Rev when they supply let's say they 65:57 you know just for purposes of discussion 66:00 they supply 10 Rev right and and for 66:05 that Rev they get so much phlogisten we're 66:08 able to you know provide so much 66:11 logistic now if they have provided a 66:15 sufficient amount of Rev for a 66:17 sufficient amount of phlogiston they can 66:19 actually perform the read on the channel 66:22 world one if they have not provided 66:26 sufficient Rev 66:30 then that Reid can't be performed so so 66:34 so imagine that the message is is a one 66:36 terabyte video right and and they've 66:42 only provided this this 10 Rev and so as 66:46 a result they don't have very much 66:48 fledged iStent so they can't actually 66:50 perform the data operation associated 66:53 with the read right and you know and so 66:56 what what should happen Utley why the 66:57 contract and let me just let me just 67:00 finish for this for the purpose of the 67:01 audience that might listen later so what 67:04 should happen is they'll get an out of 67:07 phlogiston error like a theory I'm out 67:11 of gas error 67:12 right and and then and then the contract 67:15 will roll back in the sense that that 67:17 any any data that was read from the 67:20 channel partially it will be as if it 67:23 had never been read and another contract 67:25 potentially could pick up that data and 67:27 consume it for itself um now if they 67:32 sound is bad hmm well it's if you can 67:37 hopefully read information you can leak 67:40 something in there yet no you can the 67:44 problem now it's a it's a lot safer than 67:47 it sounds because the the contract can't 67:51 can't have the read things are partial I 67:55 mean a atomic not partial so if the read 67:59 doesn't happen then the continuation 68:01 doesn't ever get a chance to execute 68:04 okay Greg I just complete the picture 68:09 and then we can then we can jump in with 68:11 with lively debate now suppose that um 68:16 that the the amount of data that was 68:20 read was much smaller than what was 68:23 allotted for in the phlogiston that was 68:27 associated to the contract so the read 68:30 succeeds so the message is read from the 68:33 channel and now you can move on to the 68:36 continuation but at that point the 68:38 continuation should not have the same 68:40 supply of phlogiston it has to have a 68:44 supply of phlogiston that is decremented by 68:47 the action the the amount that's 68:50 associated with the action of reading 68:51 and that's what Joseph is talking about 68:54 so the I seek you know theoretical 68:58 challenge with this whole theory number 69:00 one you are already defining upper 69:03 bounds and lower bounds for you know how 69:05 much it's gonna cost and my submission 69:07 that when you are tossing it then you 69:09 shouldn't get just upper bounds not 69:10 lower bills agreed so now now another 69:15 example you gave one terabyte so my 69:17 expectation would be when you are 69:18 floating a contract or pushing a 69:21 contract into Rho V M you will give them 69:23 the estimate and you charge you know 69:25 upfront and that's one way unless until 69:29 you define a contact which which would 69:31 have dynamic charging capabilities where 69:34 you know VM would charge that contract 69:37 dynamically right I mean there are two 69:38 separate kind of contact we are talking 69:40 about let's um let's kind of make a 69:46 distinction um well I guess first it 69:49 would be really helpful if um and I 69:52 guess I I mean I assumed it but reading 69:56 the etherium documentation on exactly 69:59 you know how their gas pricing works 70:02 would be would be really useful right 70:04 now the their entire approach is dynamic 70:08 it's as the contract is run with every 70:10 action every bytecode that's executed we 70:12 decrement the amount of gas by a certain 70:14 amount so what I'm what I'm saying is 70:17 it's simply that because we don't know 70:20 precisely the amount of phlogiston 70:24 that's going to be consumed at actual 70:26 runtime because there are just things 70:28 that we cannot know we take a Max bound 70:31 and upper bound statically and we charge 70:33 that upfront at runtime we decrement 70:39 from that amount per action that is 70:42 executed on the machine and then the 70:44 remaining amount is refunded or given as 70:47 and partially I think actually yeah the 70:50 remaining amount is is refunded if in 70:52 fact it doesn't require the entire 70:54 amount of phlogiston that was given 70:55 upfront so another way of stating 70:57 this is the compiler figures out a 71:01 deposit that you have to make a safety 71:03 deposit you make the safety deposit you 71:06 run your program and then you get 71:08 whatever part of the deposit back 71:11 that wasn't consumed by the program 71:14 that's correct yes exactly 71:17 does that make sense not me no I mean in 71:22 third Eagle cents yes that's one way but 71:24 I think we have already muddied the 71:27 water with dynamic content right so 71:30 suppose you don't know the size of the 71:34 read operations later on right in that 71:38 case how do you do that 71:39 I mean read that supposing in the beginning 71:42 of the service have two different kind 71:45 of contracts to buy one would be the 71:47 fixed pricing second what is the dynamic 71:49 pricing right no no it's it's not fixing 71:53 both of them together I think this is a 71:55 misunderstanding we're not talking about 71:56 there's no dynamic pricing here what 72:00 would starting the night charge no no 72:04 the dynamism comes from just 72:06 decrementing the rev that's being 72:08 consumed or the phlogiston that's being 72:11 consumed while the contract executes 72:13 that's the only form of dynamism the 72:16 pricing itself the dynamics is the 72:19 pricing is a market thing right the the 72:22 dynamics of decrementing the rep has to 72:25 do with allowing the contract to execute 72:26 as the contract is executing the pricing 72:29 is not being adjusted as the contract is 72:32 executing the only things being adjusted 72:34 is how much phlogiston has been consumed 72:37 in executing each statement oh yeah yeah 72:43 how would the estimates not be a almost 72:46 infinite so if it depends on the size of 72:48 an input which could be infinite I'm 72:51 thinking would the estimates come out as 72:54 so high that it would be impractical yes 72:58 doesn't have it's a good question in 73:01 exam room yeah 73:02 so in general think about it from the 73:04 point of view of the java virtual 73:06 machine right so even though the list is 73:09 infinite 73:10 actually infinite in length the pointer 73:12 to the head of the list is not so so you 73:17 know that the read operation on reading 73:20 a list is bounded initially by getting 73:24 that first pointer and then exploration 73:28 of the remainder of the list corresponds 73:30 to additional read operations so so that 73:35 that's where you can you can you can 73:38 look at reasonable upper bounds that 73:41 make sense I mean I think I mean 73:44 conceptually I understand Greg I think I 73:47 mean my confusion started when you know 73:50 I think one of the statement I heard 73:52 done implications of the flow distance I 73:55 think conceptually I understand how this 73:57 is going to work mean I don't have any 73:59 confusion anywhere so I think it's much 74:03 more implementation detail with how we 74:05 are gonna achieve these things in our 74:08 chain that's where my confusion came 74:10 otherwise I think I'm good thank you 74:12 well we'll imagine okay so imagine that 74:15 your Rev that the amount of Rev that you 74:17 have is contained in a register so it's 74:21 you know easily accessible and very 74:23 quick in very quick to decrement and 74:26 then you know with every action that we 74:28 execute it'd be just like decrementing 74:30 any other value or incrementing the the 74:34 instruction pointer 74:35 yep that's right Marnie well respect 74:38 logistic yeah yeah my question with 74:45 respect to our chain would be you know 74:47 suppose your contract does 10 reads and 74:50 you had spent 10 revs and a completed 74:58 nine of them but the 10th one it 74:59 couldn't complete now you would get it 75:04 out of logician error but those nine 75:08 reads that were already done so you 75:10 don't have to restart the whole contract 75:12 or no well in the case that you in the 75:15 case that that you do run out of 75:17 phlogiston during runtime then the 75:20 machine would recognize halting 75:22 condition 75:23 so it would stop and then would roll 75:24 back all the changes that had been that 75:27 had been performed that's one of the 75:28 benefits that we get for free you know 75:31 by the consensus mechanism is that we 75:33 have a guaranteed past history of states 75:37 so you know pretend like none of this 75:39 ever happened in the case that you run 75:41 out of phlogiston at runtime yeah Jim you 75:45 raised a really good point that that's 75:47 related to long long-running 75:49 transactions long computations and 75:51 checkpoints so the benefit of Rholang is 75:56 that it allows you to organize your 75:57 computation in such a way that if you 76:00 know it's reasonable to commit without 76:05 having to read all ten of them then then 76:08 you can do that you can write code that 76:12 it is aware of that verses code that 76:15 needs all ten of them in order to commit 76:17 so so so the the transactional nature or 76:22 the transactional semantics of the for 76:25 comprehension allows you to set your 76:27 boundaries exactly as you want so that 76:31 you know for example if you if there are 76:34 10 parallel reads from ten parallel 76:37 channels you could write that as a for 76:39 as a join in a for comprehension and and 76:43 then all of those have to go or or or or 76:48 the continuation doesn't happen at all 76:51 alternatively you can you can arrange 76:56 your contract in a different way and 76:58 there there's several different options 76:59 within the language so that you can have 77:02 partial commits rights and and and allow 77:06 for a resumption of a state even if you 77:09 get an out of gas error so that that 77:13 gives you a considerable leeway within 77:16 the language to be aware of the fact 77:19 that there's a correlation between 77:21 resource consumption and transactional 77:24 boundaries so that's that's a that's a 77:26 really good point you raise and one that 77:29 I think is a particular strength for for 77:32 Rholang and Rchain very very good 77:36 okay we're we're way over I know 77:39 Christian I wanted to chat 77:41 I totally I think it's great this is 77:45 great conversation let's continue 77:47 continue to have this conversation I 77:49 need to jump off but I also want to 77:52 mention to the community so no one is 77:53 surprised um I'm getting married on 77:58 September on September 15th the weekend 78:01 of September 15th I'm getting married 78:02 and and where we're going to be away for 78:06 about two weeks from the middle of 78:09 September to the end of September 78:11 Mazeltov thank you thank you thank you 78:16 very much 78:17 awesome I'm very very excited yeah um my 78:22 my partner but the marriage certificate 78:27 on the blockchain uh but but but but 78:31 but you know um I would have out of 78:35 respect and because I'm really looking 78:37 forward to it I'm going to be a away 78:40 from from you know business 78:42 communications and and even though I 78:44 consider many many people here family 78:47 certainly friends but but we've been 78:50 working together for so long it feels 78:51 more like a familial relationship is 78:54 going to be important for me to unplug 78:56 for at least a couple of weeks and then 78:59 get back to things with a fresh bigger 79:04 thank you thank you for all the 79:07 congratulations well so with that with 79:11 that I'm gonna bow out um 79:16 Christian do you want to meet on skype 79:18 or kick everyone off the channel and 79:19 they have 79:23 okay cool go shinny on I stop
Streamed live on Jun 14, 2017 Rchain Community Debrief 29