IRC log started Sun Jan 2 00:00:01 2000 what in particular? [msg(TUNES)] permlog 2000.0102 hard to tell... heh "particular" is not a good word to use when making a general purpose system.. ok http://www.disflux.net/bueno/ :) nice front page... "can you use frames?" :) heh tril: so what's standing between you and the tunes hll spec? i didn't make it i don't know html well... * eihrul/#tunes clicked on to not use frames. it's not that obsolete, and I never really got any comments on it.. sorry eihrul :D you mean the hll page on tunes.org? not the hll page but anything under ~dem/tunes oh heh. you should get into "push" content... put it on the mlist. discuss it there 12:10am -:- higby [RH@tnt155.web-ster.com] has joined #tunes hello water well, what's wrong with taking a tiny lisp and arguing about how to make it into tunes? hi higby guess what... i got my modem to work when i am not root * water/#tunes would rather argue than spend the short time he has to talk with tril silently hig: cool i love the chmod command How to make lisp into tunes? hig: now if only my modem would work again :) tril: sure `you having modem problems or whatever. just something ridiculously simple hig: yes, since i upgraded to mandrake 6.5 sorry for you if it'll make you fell better water you can scold me i am on openlinux 2.3 haha good one * water/#tunes slaps fire around like the whiny wannabe-coder he is. :) :( lol that's harsh you guies get vicious especially after trying to compile this program and ending up with 25 or so errors but it felt good and it's the truth :) fire: welcome to the wonderful world of c :) HERE HERE i need a stiff drink -:- higby [RH@tnt155.web-ster.com] has left #tunes [] eh? hm. he left Tril: so how about it? Ź_Ź ok what are we willing to do to get tunes going? I don't like lisps ack typesystem *minimal* lisp scheme ph34r ;) * water/#tunes slaps fire around like the whiny wannabe-coder he is. D: minimal lisp with what typesystem? forget scheme sorry father -:- higby [root@ascend68.web-ster.com] has joined #tunes ok i am back roll the typesystem over lisp how about a minimal weakly typed object system? heh. as root ya well i have some nmap scanning to do 12:20am water: what? "roll"? treat lisp as the lll how about a decent typesystem :) lol do any of you know anything about nmap typing typing typing...trash! make a simple object system for it nmap? whose type-system is the mop the scanning program higby: no, sorry k how do i get ppls ip's in here not this chan whois my idea is to make a database that uses my data flow model /whois ./dns ya ./dns thanks np tril: what is that like? i take a relational database and remoev redundant abstractions like multiple columns and make the values out of my own typesystem i see hm that word "type" makes me feel dirty each table has 2 columns and the tables are linkd together fire: shush I call it a functional database ;) say you have some type seems a little too "structured" to me ok hm. not my style but it seems ok then each of the functions on the type is a 2-column table in the functional database fire: my style is very structured so is COBOL ;) so if you take all the functions on that type you can pretend it is a multi-column table so, a function over the integers would have as many db entries as there are integers? just put the function names as the column names interesting rows, yes fire, it says unknown command each row corresponds to one value in the type rows? oh right \dns RDBMS have table rows and columns you know? yes i know The actual values in the type aren't printable i don't think of it that way I wrote abuot that in one of my specs why values are hidden huh? why wouldn't values be printable? fire: you might want to think of type as specifying the interfaces to which the object complies They are- but only with a function that takes the value and returns some printable form but i like to be in the raw of it er it's not important not pretend it ok well, how do you specify functions, then? not important enough to let it confuse you now 12:30am so is each row like a member? you don't... at the level we're takling about the functions are defined at a meta level.. bceause the current level is all unchangeable fire: member of what the function whoa you're separating meta-level from first-order level... why? because it will help me understand the system better. i disagree i mean, i don't think that it will help you understand the system better in any way that helps tunes tril: how do i get to your project sources? anyway, just assume that I am separating the levels for now.. they get cmobined later (hopefully after we have an implementation of this) i can't get to them from your home page try the hidden ~dem/old/ symlink ah hmm then go to the hidden ~dem/old/tunes directory :) well, what's your meta-level then? haven't decided yet ack i'm still trying to get a clear picture of the first order level darn it It will be either LISP or C, does that answer the question? because i can show how your idea is very similar to mine and how that very problem (how to specify functions) relates to how i intend to specify graphs I am also trying to make a non-reflective design first yeah, well i already have one of those the arrow alpha code specifying functions, I can answer that "at the meta-level" is not an answer :) first-order level is you can use declarative style to define every pair in the function other than that you can use lambdas or whatever which are a meta level heh it looks like most of the people here have different ideas for tunes yes, but declarational style is arbitrary and incomplete (GODEL) -:- SignOff higby: #TUNES (using sirc version 2.211+4KSIRC/981227-pre0.9) and this matters beacuse... not to mention incomputable and intractable for most cases what?!? maybe you read more in than I meant maybe i did I am saying that I can define functions by saying f(blah,blah). f(blah2,blah3). just defining pairs 12:40am but i've been studying up intensely on these issues, and i think that there are some very important issues s/issues2/problems And lambdas are just a way to do a whole bunch of such declarations all at once. tril: which is like my graph spec of functions without even doing the declarations. A graph spec, suer sure yes, but you can't specify '+' with lambdas without resorting to recursion through the natural numbers and you can't specify '/' with a lambda or '-' for that matter so... use recursion then... or do you want to do it without lambdas? the problem is that you'll wind up with lots of complexity at the meta-level that will not get you any closer to tunes-capable reflection the system for specifying your functions has to be your biggest concern what's wrong with defining + like we did in scheme class, calling Increment recursively nothing with that, but defining - that way won't work we did that too or / and / and pow how / ? I'll go grab it :) whatever it was, it was only programming /, not a mathematical / there's a big difference that programmers don't get It was meant to be a mathematical definition, it was very inefficient. well, that doesn't help specify the function what do you mean "specify"? don't you just need a way to get the result, given the input? because it doesn't allow the system to relate rationals to integers no specify means understanding otherwise, you don't get "useful", you get "expedient" :) BRB 12:50am back water isn't talking about lambda closures understanding?! whose understanding anyone's I'm sure there are people who are really good at specifying functions Whatever language they use to do it, they should be able to use that well, that argument is one that the tunes members haven't really agreed with geez I'm not raelly concerned with it you missed the point Yup i've been trying to figure out how to specify objects as well there are all sorts of aspects to objects that are not captured by calculable functions some insight would be nice in real life, every model that a computer can manage will diverge at some point from the actual dynamic system think about the word "specify" what does it mean. It means to get more specific. To specify a function all you do is start with the set of all possible associations ebtween members of the two types then you use some langauge , ANY LANGUAGE that allows you to describe which associations you want and don't want. yeah set theory but the language doesn't have to be lambda-calc I'm not an expert on those specification languages.. nor do I have time to be you assume lambda-calc implicitly I didn't say it had to be why do you think so? well, how else would you specify functions? maybe it's the only one I know s/you/*you* what if you don't have all the information on what a function should do? what if you can't specify the domain precisely? the first question is saying what if I can't specify the function precisely. no, it's saying that the *goal* of the function can't be known then clearly I want to reason abuot the function in the abstract. the computer isnt expected to implement it. for example: how do you analyze experimental data? I don't think of functions as "doing" or "having a goal" they're just static mappings oh wait! abi agt hmmm... agt is Around Goedel's Theorem at http://www.ltn.lv/~podnieks/ that kind of thing is what my professor specializes in.. we learned a lot about it in the AI class he taught Inductive Logic Programming 01:00am yeah, and it's not the world a program writes a program and then refines the body of the program as new data comes in that's not all there is to it I won't object for a given physical situation, there can be many different conflicting yet applicable computational models and prediction in most cases isn't as valuable as understanding ... have you read agt anyway? I'm not sure you have made your point no i think you should at least look at it I'm looking at the page Is this a book with the entire contents online? If so, cool. ok, a long thesis it actually was a book in russian in 91 i've also been looking at hyperset theory it's where you allow sets and classes that are members of themselves is that what you mentioned the other day, yes it is sounds interesting yes, especialy for reflection hcf came up with some links, hold on what makes them better though? "better"? er it's a more expressive theory why would yu want hyper sets oh well it should be allowed "allowed"? er im not one for words tonight possible i guess in the language 01:10am the problem is that some important axioms have to get shifted around but it allows graphs with loops and identity arrows to be faithfully encoded by sets and/or classes :) http://www.cs.bilkent.edu.tr/~akman/jour-papers/air/air.html it's a related paper from an ai journal so, what's your non reflective system that paper on platonism is really good water huh? im glad you pointed it out earlier my arrows code um thx with the platonism part that is heh im really not well tonight there's also a paper about unifying information theory, if you're interested YES! heh i would love to read it i thought i was the only one http://splendor.unibe.ch/Activities/UCI/Proceed/FIS96PRC.htm if the word "unify" is there i'll read it if you're not familiar with shannon or formal info systems, it might not be easy to read yes that might be a problem it's great that i can carry around so many papers on my handheld im sure a search on altavista will solve my problem though yes, there's a copy of shannon's information theory on the web, iirc and formal info systems are just oo programming systems the next thing I'm going to do is solidify my partial evaluation model. then back to the tough problem of a type definition language darn it, i just think you're working on the same problems i am are we? i believe so we're concentrating on different aspects of it, though i'm not sure if we have exactly the same goal my goal is to get some code soon so the CS dept doesn't force me to take their C++ intensive degree program brb C++, X86 ASM and Win32 why would your code help that? that's why i dropped out 01:20am (in a manner of speaking) if I just write incoherent papers and reams of notes, I won't be approved for a custom degree but if I write some code, I can say, this is my degree, to work on this program they really let you do that? heh. how much time do you have? :) no. I have to take classes that are related to my work, and show how they all relate with some central project. time? A year, minimum if I do my own thing. Probably 2 or more if I have to take a bunch of boring CS classes -:- SignOff fire: #TUNES (Ping timeout for fire[209.68.229.40]) hm i think you use the word "partial evaluator" in your spec a bit oddly I use it to mean specializing of functions if a function is specialized all the way, it has all its arguments known and there's nothing left to do because we have the result but how does speciallizing functions mean "partial evaluation"? that's not the same thing, imo of the function perhaps? well, if you have f(x,y) and you pick a specific value of y, you get f'(x) = f(x,VALUE) right, but the domain of x's is not a subset of the domain of (x,y)'s ahh, so closer to memoisation than what partial evaluation usually means that's why i say it's odd but it's isomorphic to a subset of it. the set of (x,VALUE) but that makes sense I guess :) right, but it's still not strictly the subtype it requires that isomorphism function to relate the two that's true i just think you should explain that -:- SignOff hcf: #TUNES (Read error to hcf[me-portland-us924.javanet.com]: Connection reset by peer) -:- hcf [nef@me-portland-us203.javanet.com] has joined #tunes i don't think that there's something wrong with the notion, just that you should be clear well, the standard kind is also included if I wanted to specialize with subtypes, I mean what's that called? partial evaluation? sigh... no, specialization with strict subtyping according to you guys, that's called "partial evaluation", and what I'm doing is slightly different; 01:30am no it's not I probably use "partial evaluation" where I just mean "specialization", is that the confusion? under 2.3 of the spec, you list Specialization (Partial Evaluation) what is the distinction then? Keep in mind I've never actually used a partial evaluator. specialization by subtyping is not partial eval of the function, it's partial eval of the set or something like that So, you're saying my terminology is not clear, but you dno't know what it should be? i'm saying you should split the two ideas I'm asking you to tell me what partial evaluation is hm "what it is" is evaluating the function to a further lambda-context I'm basically going off of the Glossary definition subtyping is different from that yeah, and it's not the world oops ignore that * water/#tunes checks the glossary A metaprogramming technique that consists in transforming programs into simpler programs specialized for execution in an environment where some information is known as compared to a generic environment. Typically, a function that takes many arguments is partially evaluated considering as known the values of some of the arguments (the static arguments) while some other arguments are still to be determined (the dynamic arguments) um. was that necessary? no. sorry. It didn't look so long on my screen. look... metaprogramming a program into a simpler program isn't the same as restricting the domian by subtyping the latter is a particular case of the former i disagree then what? in order to partially-evaluate, you have to apply some function to take a subtype doesn't involve a function in the normal sense I'm assuming that we have selected a particular function of course. ok 01:40am When you said "subtyping a domain" obviously a function is implied since "domain" has to be domain of some function ok, thats not why let me just give the example if I have a function, and then I find some subtype of its domain, * water/#tunes checks the names Fare references in the glossary from that, a specialized function can be produced. sure What's that called? domain-restriction, i guess heh. here i go again, downloading papers so how does it relate to Partial Evaluation it doesn't precisely relate to it: that's my point and I don't understand why because if you're going to use type theory, then you have to understand that, for instance, the real number line is not a subtype of the real number plane A further lambda-context you said, but I say that is just the lambda context f(x,y) with y replaced with 1 it's how you arrive at a new lambda-context that is the difference one way, you take a subtype. the other way, you factor out a "dimension" from a type which we already said could be done by first taking a subtype, then showing an isomorphism . So what's the problem, just show lots of isomorphisms. (darn it, if only someone would port ghostview to the casio :) heh s/someone/you/ it was a joke, ghostview would never fit on a casio I like this factoring out a dimension idea.. you think I have two concepts really confused? well, it's not subtyping it's certainly specialization