IRC log started Sun Jul 4 00:00:00 1999 [msg(TUNES)] permlog 1999.0704 -:- water [water@ppp-tnt-127.tscnet.net] has joined #tunes -:- SignOff _QZ: #TUNES (BRiX [http://www.qzx.com/brix] :: sleep) -:- sMike [user5739@ppp09750.01019freenet.de] has joined #Tunes -:- sMike [user5739@ppp09750.01019freenet.de] has left #Tunes [] -:- BaKneeZzz [chrisg@glider.chrisworld.co.za] has joined #Tunes Lo There... 03:20am -:- SignOff BaKneeZzz: #TUNES (Segmentation fault : KVirc 0.9.0 by Szymon 'Pragma@ircnet' Stefanek ) -:- SignOff water: #TUNES (Ping timeout for water[ppp-tnt-127.tscnet.net]) -:- FareAway is now known as Fare -:- smkl [sami@MCCXXVIII.rdyn.saunalahti.fi] has joined #tunes -:- water [water@ppp-tnt-250.tscnet.net] has joined #tunes -:- SignOff smkl: #TUNES (smkl has no reason) -:- smkl [sami@MCCXXVIII.rdyn.saunalahti.fi] has joined #tunes wb hello gakuk yablutsk! :) you gaklosmontic! hehe * water/#tunes reads a positive review of the "south park movie" on slashdot any topic to discuss? hum. Weather? hehe. foggy are you in a good mood, today? i hope so :) 08:50am hum. The chair is more comfy, here it'll take a lot to change my reputation * water/#tunes lives with many angry and conflicting people. they seem to be affecting him. :( that's ok :) I seldom get irritated in e-chat it certainly explains the difference in my net personality since i was last on 4 years ago after all, there's nothing to give/take but information... yes speaking of which... what's on your mind? perhaps we could teach each other about a few subjects here or not, i guess 09:00am my ISP! fare, your isp is crappy hehe wb water: met a good friend yesterday whom? Basile.Starynkevitch@cea.fr ahh! how did you meet? physically he lives in the suburbs of Paris... ok you discussed... ? sure he'd like to do AI stuff and regrets that AI and formal methods communities, whereas technically complementary, are mostly despising each other as do i so do i 09:10am "went something like this. (Laser hologram sputters, goes out.) [Army General] "Damn, it's Windows 98!" (Drags Bill Gates in). [AG] "I thought you said Windows 98 was a faster easier way to the internet... (etc)" [Bill Gates] "But Windows 98 IS a faster and" !!BLAM!! (Bill falls to the floor dead.) " a good part from the south park movie faster than what? than Windows 2000! hehe With MICROS~1, every version is even faster than the next one! unfortunately, i am still in transition to open-source a hopeless addict :( i keep wanting to get beos or to wait for a powerpc g4 machine but they're mutually exclusive goals :( bleh 09:20am go linux, now! i have it it's on 50% of my hard drive great! now, wipe the other 50% :) hehe i guess i just have to get around to putting my back-up data on a ditto tape that would allow me to wipe it of course, my ditto drive is on the ship well, if there's nothing to discuss for a while, i'll see you later bye -:- SignOff water: #TUNES (The Tao has left the building!) 09:30am -:- tcn [tcn@cci-209150250111.clarityconnect.net] has joined #tunes >>> tcn [tcn@cci-209150250111.clarityconnect.net] requested PING 931106339 219736 from #tunes hey ya'll 09:40am abi: time? i guess time is 09:41:54 1999 -:- SignOff tcn: #TUNES (tcn has no reason) -:- HickServ [thrustit@209-68-229-167.dialup.cust.tfb.com] has joined #tunes hola a todos 09:50am francois: eres uhhh...awake? :) 10:00am dfgdfg * Fare/#Tunes is back estoy re'veille' que' tal? -:- hcf [nef@me-portland-us240.javanet.com] has joined #tunes 10:10am -:- eStormy [stormy@rain.futuresouth.com] has joined #Tunes -:- jdl [jiml@ultra1.inconnect.com] has joined #tunes Hello all hello jdl :) I wanted to talk to Fare, but it looks like he's gone... brb back jdl! Fare! gakuk! ;) Hey, I wanted to ask you some questions... ...if you've got a minute sure The biggest thing I wanted to talk to you about was the questions I brought in my last email ("Challenge my assumptions") Your assumptions are quite flawed. as says the other guy "they plain suck" :) The other question was if you knew a more appropriate forum for Prism discussions... although Prism could be adapted to TUNES, that's not my goal... 10:20am prism@prism.org? Okay, that's why I wanted to talk to you. Why are they flawed? orders@prism.com? :) Not enough audience yet to warrant a dedicated forum. 1) is not a useful property per se (Turing tar-pit) 2) is rubbish 3) is dubious okay, now back up your statements. Or, what would you recommend that would also meet my goals? you might wanna read "Type-Logical Semantics" by B. Carpenter (sp?) Is it on the net? www.amazon.com? he meant free I'm not buying any more books at this time... I'm between jobs, need to conserve. Besides, I have too many half-read books already. " usually ships in 24 hours" the seemingly official page for it is at http://www.COLLOQUIAL.COM/carp/books/tls/ eStormy: I'm not the publisher To answer your statements: 1) Not useful by itself, no. But necessary for me to meet goal #1 2) Why? 3) Why? 1) the property is necessary but not sufficient. your model does not have all necessary properties 2) we could have a LONG discussion on "intuition" let me just say that your data types are certainly not "intuitive" in any helpful sense -:- _jdl [jiml@ultra1.inconnect.com] has joined #tunes terribly sorry old chap 3) is meaningless unless somehow formalized; and if you do formalize, you'll see that your metamodels keep too many things intensional to be useful -- there are too many things you CANNOT express extensionally with your poor basic blocks my brother had to look some stuff up 10:30am fare which university do you attend? <_jdl> I'll address 3) first, since it's more objective... give me an example of something I cannot express HS: www.ENST.fr (formerly www.essi.fr and www.unice.fr, formerly www.ens.fr) (the most prestigious being www.ens.fr) oh well i met a foreign exchange student from france today jdl: do you master lambda-calculus? he goes to sabon or however the hell you spell it <_jdl> Fare: master, no, understand the basics, yes if not, you should REALLY have a look at it before you explore metamodels any deeper fare: is there anything *good* online on lambda calculus? jdl: most intelligent and cultivated CS researchers have developped _lots_ of metamodels already. e.g. free Having original ideas is no excuse for not acquiring basic common knowledge. <_jdl> Fare: sure... but you still haven't given me an example. remember newton: "if I could see that far, it was by standing on the shoulders of giants" you may be tall, if you stand on the ground, you won't see much far. masterfully said fare <_jdl> Fare, I don't claim to have invented metamodels. Now how about that example? jdl: rewrite logic, lambda-calculus, denotational semantics, natural semantics, operational semantics... jdl: how do you express these in Prism? In natural language? If that's it, you're not serious. <_jdl> Fare: One thing that came out in discussions with Tril & Brian yesterday was that people seem to misunderstanding difference between how I use models and metamodels: jdl: also first-order logic and model theory; its classical expression; its intuitionnist expressions; etc category theory, etc abi: first-order logic? hickserv: wish i knew <_jdl> Models are defined only with bit,map,stream and are contained within the Prism environment :( jdl: maybe I missed something deep about Prism. Maybe you did. <_jdl> Metamodels are defined with natural language and are NOT contained within the Prism environment. They are used by humans to manually write programs which manipulate Prism models in the environment. -:- SignOff jdl: #TUNES (Leaving) -:- _jdl is now known as jdl jdl: bit, map, and stream, are ridiculously rudimentary as compared as what you get with data-structures within existing formal systems jdl: it's hardly better than Turing-machines; at least Turing-machines were purposedly made ridiculously minimal. * HickServ/#tunes has rug burns from wrestling with his brother Perhaps, but let me finish arguing point 3 * HickServ/#tunes is desprerately looking for attention HS: for classical first-order logic, get Jech, "Set Theory" * HickServ/#tunes is like a dog oh * Fare/#Tunes throws a bone at HS bark Since metamodels are defined with natural language, I think I can create a metamodel for, say, lambda calculus. In fact, that's one of the first 'real' metamodels I intend to create. like universal and existantial logic? if that's "natural language", it's as if you haven't built anything 10:40am I needn't Prism to build things in natural language If I'm to use nat. lang. somewhere, Prism becomes part of the problem set, not of the solution set. or like union and intersection? As a simpler example, I've created a metamodel for integers. The metamodel is unbounded, but of course in reality I'm limited to storage space. Fare: The C language is a metamodel which is defined with natural language and BNF. Does that mean C is useless? jdl: no. That means you had to use nat.lang. *ONCE* so as not to have to use it anymore jdl: if you can't understand that, you're hopeless. i see no use for natural language in programming but it's obviously very important for ai Fare: metamodels are analogous to language definitions. Models are analogous to source code. (in Prism) if your metamodels are nat.lang., then Prism brings NOTHING, ZILCH, ZERO, to programming without Prism oh, yes, it brings Yet Another Burden Layer I don't know what you mean by "programming without Prism" Perhaps you mean that I'm not contributing to the discipline of CS, just creating another programming environment? jdl: not even not even what? not even a programming environment. uh i beg to differ me too :) and one that does LESS than already do things like CENTAUR, POPLOG, or even EMACS! i think it's a great idea that you're trying to create something new or better even with fare's critisism fare: ahem...constructive critisism trying is good. trying with a blindfold is bad. Fare: I'm trying to create something that is practical & useful to software engineers. I'm NOT trying for the lofty goals that the people on the TUNES project are, because I want to have results in the foreseeable future. I haven't been careful attention, but isn't fare criticizing jdl for creating a programming language that, if not used, isn't useful? isn't is true for all programming languages, in that you have to use it to get any benefit from it? 10:50am eS: that, even if used, doesn't bring anything, and hence just adds to the confusion * eStormy/#Tunes goes back to reading a mailing list, having wowed the channel with his wisdom. eS: if he ain't using any formal method, and resorts to nat.lang. for any and all semantics, his language is useless than a turing machine, and I'd rather use C. Fare: That's what I don't understand. I've made my goals very clear in several posts and articles. I backed them up extensively in _The Will and the Word_. How can you say Prism doesn't bring anything? It's AT LEAST as useful as C, and jdl: how is what you're doing *any* useful, as compared to existing stuff like OCAML, Coq, CENTAUR, NuPRL, SML/NJ, Mercury, Haskell, etc, etc, etc, etc no, it's not at least as useful as C. Part of the utility of C is being standard. potentially more so. C resorts to natural language to define semantics... what so wrong about that? Maybe it's not ideal, but like I said, I want to get this thing done. If you bring no expressiveness, and lower the standardness, you're hurting the community, not helping it C resorts to nat.lang. ONCE AND FOR ALL. In the standard. Fare: is it possible that, while you may not find it of any personal use, others may find it a valuable contribution to computing? I will take a non-turing machine natural language over c any day. Fare: Prism is different than existing languages in that it focuses on interoperability. No environment does this seamlessly or extensibly. Afterwards, machines can run C programs without having to prompt the operator for "hey, what does this mean?" at every other instruction Fare: Prism resorts to nat.lang. ONCE AND FOR ALL in each metamodel standard. PROGRAMS do not use nat.lang. IN EACH METAMODEL STANDARD hence, it's useless * eStormy/#Tunes decides he doesn't understand the conversation and goes back to his mailing list. the ANS/IEEE/ISO efforts are much further than you are at defining language standards... Fare: whats CENTAUR? not to talk about the academic research community Fare: If I only allowed one metamodel, and defined it up front, then it would be useful? It's only because I allow extension that Prism is useless? hcf: look for project CROAP at sophia.inria.fr or so a system is only as useful as what it can express. If you must resort to external means to be useful, then you're not being useful *by using the system*, but *despite using the system* see DOS "hey, I can do everything with DOS" "no, not *with* DOS, but *around* DOS" Fare: It's clear that you disagree with my goals. Fair enough... but can you criticize Prism within the bounds of what it is? Fare: cant connect to http://sophia.inria.fr what's left to criticize, once you realize the bounds? Specifically, given that metamodels ARE natural language and ARE NOT going to change, is it easy to create new metamodels by using bit,map,stream (and including other metamodels) 11:00am jdl: you don't help. You just *get in the way* * jdl/#tunes shrugs Okay, where should I go? Is there a forum/newsgroup devoted to language discussions that you're aware of? python is utterly useless for boot loaders...now that we have defined the bounds, it's clear that python is a worthless language. no point in discussing scripting, which python just might be useful for. hcf: croap was split in 3. hahaha...talkk to bineng about that I thing those maintaining CENTAUR are project LEMME python is his daddy * eStormy/#Tunes likes python too. though not for boot loaders. comp.lang.misc jdl: what advantage is there in writing a prism spec, as compared to writing a compiler in C or targetting as people do now? s/in C or targetting/targetting asm or C/ not only isn't there ANY advantage; you're only getting in the way of efficiency * eStormy/#Tunes idly wonders what advantage is there in writing a tunes spec, as compared to writing a gui for linux or targetting as people do now? how are Bits, Maps, and Streams a progress as compared to BNF? Fare: Prism is designed to allow seamless interoperability. Prism specs can take advantage of that. You can also write compilers which translate models to other metamodels rather than machine code, which is easier... only one compiler worries BNF is *much* more versatile! about machine-code optomization. Fare: ah... now we're getting to some good discussion. Grammars are technically mastered, more powerful than your petty particular-case constructors, have a clean mathematical formulation that has been independently rediscovered in many fields of study, confirming its interest. I chose not to use BNF/EBNF for several reasons: 1) They're limited to text-based languages look beyond BNF -- look at the concept of *Grammar* 1) FALSE Grammars are NOT text-based Grammars are Grammars jdl: see on deja.com what Erik Naggum of comp.lang.lisp says about early optimization... 2) Each atomic unit (a character) does not represent an atomic concept. Multiple characters must be considered together to form an atomic concept. That makes understanding the source code more difficult 2) FALSE the concept of grammar is self-standing. abi: lemme is at http://www-sop.inria.fr/lemme/lemme-eng.html Just because you have textstream<->grammar mappings doesn't mean grammars are text streams! abi: centaur is a generic interactive environment generator at http://www-sop.inria.fr/croap/centaur/centaur.html hcf: thanks! 1r) I have to choose some way to represent the grammar; there's no universal format that encompasses both text-based and visual languages that I'm aware of. Fare: np ;) 1r) FALSE EXCUSE you may well standardize on grammars without having to priviledge a one format Fare: ur gonna look at lemme's pub list right? there exist (infinitely) *many* universal formats to encode grammars -:- SignOff eStormy: #TUNES (eStormy has a very very good reason) 11:10am Fare: Have you looked at the research on visual grammars? bit,map&stream are a GIANT step back, as compared to grammars jdl: I don't care about visual foo jdl: I boycott anything "visual", "3D" or "multimedia" Fare: I do. Visual languages are often a more intuitive way to represent information. For example, road maps are much more accessible when represented with a visual language, rather than a textual one. Fare: That's why I chose not to use a text-based grammar for my internal format, or a format based on text (such as syntax parse trees) programming is non-intuitive Fare: The research on visual languages is still in its infancy... and not compatible with text languages, anyway. So I had to create a format which was language-neutral. it's the victory of reason over intuition the concept of grammars is language-neutral; grammars can have "visual" or "olfactive" interfaces, if you care. Fare: That's what I'm trying to change. "Programming" a word processor IS fairly intuitive. "Programming" your car IS fairly intuitive. Programming your phone to call your friend IS intuitive. abi: croap is at http://www-sop.inria.fr/croap/CROAP-eng.html there is no reason why you couldn't interface them to people through their sexual organs, if they so desire the importance is the internal structure of the information, not the interface jdl: programming a car is non-intuitive. It takes tens or hours of courses to master a car. Re: grammars being language-neutral... the grammars represented by BNF certainly aren't. They map naturally to text, unnaturally to other representations. Fare: re car: compared to hundreds of hours to master C. jdl: 1) BNF is not about grammars, but precisely about mapping grammars to text. jdl: and a few hours to master Scheme 2) you can have self-standing grammars in Coq (called inductive types) Fare: *master* Scheme? I would disagree. (or OCAML) fare: you are in sync with hackerdom are you not? HS: I'm not sure I follow your question 11:20am well are you familier with black magic, black art or vodoo programming? Fare: okay, suppose I use grammars as my internal format. First, which kind of grammar should I use? Second, how do I represent the grammar's primitives? s/vodoo/voodoo/ black magic is an Apple ][ game I loved. no no jdl: you use types in OCAML or in Coq, or another similar language. it's supposed to be a programming method that is not documented but no one knows how it works http://foldoc.doc.ic.ac.uk/foldoc/foldoc.cgi?voodoo+programming HS: Voodoo programming is bad practice. i just want to know what it is Fare: I haven't looked at Coq yet, although it is on my list of things to do. HS: See the Jargon file. * Fare/#Tunes uses dict abi, Jargon for voodoo programming? wish i knew, jdl drat jdl: not 'jargon for', 'foldoc for' jdl: where the hell do you think i found it? HS: So now you know what it is :) it's definition is rather obscure it doesn't really tell me what it is that url you gave me was the jargon file Voodoo programming is the use of special language/machine tricks to obtain an obscure result or miniscule increase in performance. 99.9999% of the time the effort in maintaining such "wizardly" programming far outweighs the benefits of the code. dict "voodoo programming" oh oic ok well i'll be right back jdl: that's not the jargon definition jdl: what you're describing is just C programming. (or worse, C++ programming) * jdl/#tunes grins Fare: C isn't bad... it's a straightforward mapping to the machine.. low level, but not too hard to understand IMO. C++ on the other hand is a gift from Hell. C *isn't* a straightforward mapping to the machine. Hence, C-- 11:30am Fare: I'm looking at Coq... is Coq a language? If so, where is it? All I see is the "Coq proof project" go ahead s/project/assistant righto for a language, see OCAML Okay, where do I find the info on grammars in Coq? (Coq is written in OCAML, and can extract pure OCAML code from proofs) Coq manuals? Hum, hard to read. Especially the grammars part. But if you consider Coq inductive types as self-standing grammars, then just read the Coq tutorial Coq "grammars" being just mappings from text syntax to Coq objects and meta-objects. okay, looking at that (CIT) oh, and Coq *is* great, but it's not the be-all end-all of proof systems CIT? Coq inductive types ;) Okay, so looking at Coq's inductive proofs, it seems that the primitives of the system are sets and logic uh fare so what are you telling me? jdl didn 't telll me what vodoo programming is? 11:40am HS: Yes, I got confused with black magic. no. Primitives are typed lambda-calculus enriched with inductive definitions. HS: go read your jargon file oic damit fare i did like three times i first read about it there HS: Voodoo programming is doing something and hoping it will work without any understanding of if it will or why it did (if it does work) s/any understanding/any real understanding/ oh I've seen "voodoo debugging" quite a bit when multithreading was involved :) hey fare are you awake buddy? i've been trying to msg you so? Got to go... see yall later. Thanks for the discussion, Fare -:- SignOff jdl: #TUNES (Leaving) 11:50am continue good fare well, what can't you do? have you read SICP? or another Scheme tutorial? even material on www.schemers.org? i understand some *basic* scheme i read a tutorial so you know how to make a closure? don't tell me in scheme but just give me ideas it can be language independent fare: no you know how to emulate objects with closures? you should definitely get SICP or another book that's more than a hands on tutorial to basicest scheme I'm sure there's stuff on www.schemers.org well, come back when you have actually tried. or have questions. hmmm how about you tell me in C :) I mean, I could write the thing, but then it wouludn't be you! no no i don't want you to write it HickServ: then get a scheme equiv to lex/yacc s/then// but i need to know some stuff on how to get the important stuff into the backend 12:00pm why don't we go back to where we were before when you were so enthusiastic to help me let's see first you must write the lexer well fare i've already written it in C well, rewrite it, in Scheme. if you have no idea, then just make it a straightforward port. HS: you have a main (define) ok then let's imagine i already ported it; now, it takes Word and runs it to (parse Word) in that define, you define helper functions HS: then what? then what is my specialization? well now i must define (parse Word) and make it parse Word is a 34 pair list containing either words or parentheses er s/words/word i.e. a word can only be 32 bytes long why such limitation, ? well i gotta limit it somewhere with Scheme, no need to (even w/ C, there are GNU obstack) or with mmap, similarly 12:10pm you mean like append or w/ C, malloc? w/ C, GNU obstack i'm not familier with that w/ Scheme, either string-append or you own obstack implementation HS: time to learn fare: indeed let's just leave it at malloc for now ok then it takes Word which is a string without whitespace now it must parse it mind you it's SEXP don't tell me your isp fux0red you again 12:20pm malloc(howmuch)? no. the tokenizer takes a sequence of tokens each token is already a parsed atom, or a special token special token, e.g. "(" ok so what i would make the prelexer look at the first character or so of Word and then decide what kind of token it is ? 12:30pm