IRC log started Sat May 20 00:00:01 2000 [msg(TUNES)] permlog 2000.0520 -:- water [water@tnt-10-214.tscnet.net] has joined #tunes re all -:- water has changed the topic on channel #tunes to: TUNES, Free Reflective Computing System: http://www.tunes.org/ || Slate Language: http://www.tunes.org/~water/slate-home.html 12:50am -:- scipient [scipient@ip254.charlotte17.nc.pub-ip.psi.net] has joined #Tunes hi scipient hello. there seem to be a lot more users than conversation, eh yeah but usually the lurker count is higher well you shagged me out of the carpet is there something i can do for you? heh um, i dont know.. not specifically. just checking this place out ok i assume you came after seeing the website just discoverd the tunes web site. it inspired me to download an irc client. this is essentially my first time on irc except for a desperate attempt to get help on a linux install last year heh ok you are the fellow behind arrow / slate, no? well i'll gladly answer questions or explain things for you yes, that's me "explain" sounds like a monumental task in light of the scope of tunes yes in a lot of ways it can be it was an amusing twist of events to discover the site, for me, since i've been thinking a lot over the last few months about something more or less just like tunes the people who make it here tend to have ideas like tunes i've found yes. (like myself, actually) right ;) ... i have by no means fully read the site yet. too tired to concentrate that hard do you have a page that describes your idea? no hm ok it's more a loose collection of ideas / reactions to the daily crap i deal with from windows ah well hopefully we can help you organize them i've been thinking about trying to develop a sort of universal user agent hm made of an unbounded set of com/corba objects (depending on your platform) which are glued together with scripting languages alright... sounds decent enough mainly so every user can customize their own GUI and use only the parts they want and customize the interaction of the parts actually there's a channel for a project much like that tunes seems to be what you get if you keep thinking like that and apply it to everything #dolphin (though no one's there right now) no tunes is more than that, actually dolphin -- was that one of the OSes on the review page? hm yes, i see that tunes is much more actually this particular dolphin is still in development it incorporates some other ideas i had, plus things i never thought of but it is a tiny kernel that bootstraps an ORB would it be a "microkernel" architecture ? are you familiar with Lisp or Smalltalk? 02:10am similar notion, yes nope. please, enlighten me, if you can in a few sentences hm a few sentences will not do, but i'll try i noticed on the OS review that Lisp seemed to carry a peculiar mystique of esteem lisp is the programming language based on CONS cells for its syntax AND data-structures, which allows it to treat programs as data yes, because lisp is good at reflection hmm. what are CONS cells? a cons cell is the basic unit of a binary tree my_cons_cell=(a b) so the Lisp program can be "graphed" in a binary tree, maybe? exactly and then you can collapse the whole tree into a node of another lisp program among other things, yes smalltalk demonstrates another language principle that it's not just object-oriented, but recursively based on the same level of abstraction i was wondering, maybe the appropriate language for tunes will be purely visual.. some kind of 2d or 3d environment where you use a mouse to create and connect objects together, and you can sort of watch the program flow through all the nodes yes, but other projects are also working on that a bit each node could perform decisions, operations, communications with other nodes, etc. have you heard of VPL's? abi: vpl oops the info-bot is gone, sorry ?? no, not VPL or abi: vpl info-bot ? yes, "abi" is our usual infobot ok. generates blurbs on demand? she carries a dictionary of sentences about things, usually with URL's i'm sorry i missed her yeah she's rather useful anywa anyway, VPL's are visual programming languages ok. i played with LabView in college ah yes ever heard of Prograph? i think it would be a trip for someone to try to communicate the source code of a VPL program in words it's not impossible no, no Prograph. i do not come from a comp. sci background but yes it's rather difficult heh if it's not impossible, then of course you could create a representation of the language in a textfile prograph is a language that's been commercial since the late '80s i read an article by some XML advocate that suggested there would be a standardized XML representation for code of course it only ran on Macs until recently, so few knew about it standards are ways people make up for the lameness of the underlying tools and all our favorite languages would compile/decompile to or from that XML DTD what is Prograph used for? actually, you should check out Squeak it's just a good IDE http://www.squeak.org/ Squeak. I think I saw that on the OS review too ? yeah it's a portable smalltalk written in itself interesting. it's actually an OS also when you link it to a bootloader 02:20am so is Oberon, but oberon is less impressive what could you do with it? with squeak? with squeak one hell of a lot of things i mean, could you have the same functionality you get with a typical desktop OS ? it has a web server, a mail server, foreign function bindings, 3d rendering, ... it's own native gui's built from smalltalk, a music engine the list goes on impressive. what does tunes do that squeak doesn't? it's not the fastest thing in the world right now because the cross-platform JIT framework is not integrated, but that will come soon heh a whole lot squeak is missing reflection in a big way i am just trying to get a feel for the commonalities, so perhaps you should take the complement of my question phase space partial-evaluation and migration in squeak are pretty much completely absent other than the language features hm ok could you loosely describe reflection as the ability of an OS to emulate itself ? well that's one kind of reflection say, system-level causal reflection i admittedly scanned some of the tunes pages rather quickly reflection is also the ability to modify the programming language used dynamically heh which sort of implies that the os understands itself and to dive into the implementation and optimize it for the currently-interesting case well there are many degrees to it it almost sounds like a gateway into AI there's also reification grrrrr sorry if i offended you don't even mention that word in connection with tunes that's the last thing we need is association with the broken ideas of ai ok. i suppose that is one hell of a loaded buzzword hm broken sentence above s/is/;/ i blame it all on those sci-fi novels i read when i was a teenager hm s/;/!/ heh the only ai-related scifi i ever read before i had a clue was 2001 (thank god) i liked neuromancer a lot neuromancer was good, but not unless you understood what was bs and what wasn't i usually have a good eye for that s/not unless/only if/ bleh i'm getting tired most people don't read their sci fi as critically as you, i think what time zone are you in anyway? no they don't, but then i'm an obsessed person PST well you've got 3 hrs on me 02:30am so, suppose I was an irritating internet journalist and I said: Mr. Water, what's wrong with AI? (and you felt compelled to answer politely) heh * water/#tunes hmms you want a soundbite or something more substantial? a little more substantial did you read geb? in other words, what makes tunes realistic and AI bullshit? sorry, godel escher bach i read about half of it i'm sorry are you sorry for the half i read or the half i didn't? that thing bakes people's minds it seems anyway about your question yeah, he lost me when he started describing the human brain as a finite set of algorithms (including the 'me' algorithm) AI consists of people attempting to encode self-sustainable intelligence on a machine with the implication that English was some kind of programming language total nonsense yes. i doubt you can make intelligence out of any number of 1s and 0s anyway, it's impossible for computers to reason in the sense they desire, because symbolic manipulation cannot be bound to the real world in any formalizable way i think if we ever wind up with anything close to AI, it won't be the result of any encoding process well i don't even think things like neural networks will ever get it either more the result of a quantum system evolving to an incredibly high degree of complexity sorry, i feel that this is rambling granted. but i think your statement about the impossibility of binding symbolic manipulation to the real world is an excellent summary ok . keep in mind of course it works when the computer is only seen as an information intermediary (this ties in to my arrow ideas) can you elaborate on that? what was the antecedent of "it works" binding symbolic manipulation to the real world 02:40am there's a book that gets into the philosophy that supports this and elaborates well on the point ? "Understanding Computers and Cognition" ok don't worry, it's a short and cheap book reflection is the ability to modify the programming language dynamically.. this is something the OS would initiate itself? hm sort of keep in mind the os is *programmed* to do things so there would be a well-defined decision procedure for it to do that so meta-programming is what does this (modify the language) if you gave it one, yes how else? you could also give it a non-deterministic algorithm you mean, a sort of mutative darwinistic process? * water/#tunes sighs not necessarily, but that's an example, yes sorry. the OS might experiment with modifications of the programming languages which satisfy certain criteria, and benchmark them for performance, and remember the improvements? yes we like to just call it a "system" instead of "operating system" ok. since it's operating parts of itself the positivistic noun-verb view behind "operate" is too blurry to enforce and we don't even want it heh a legacy term you tend to get stuck with, i suppose seriously, tunes will likely have no kernel per se -:- smoke [smoke@16dyn143.delft.casema.net] has joined #tunes that is an interesting part. presumably, there will be some assembly code on your x86 system that will bootstrap the whole thing hi smoke hi hm of course everything eventually gets compiled, and yes of course assembly has to be at the bootloader section but that doesn't mean we can't have the bootloader be expressed in a high-level language yes. and change it if we choose now, would the system retain a memory of the fact that it was bootstrapped, or would it forget that aspect of its state information? 02:50am huh? which bootstrapping? from hardware power-on? well, it seems to me that a typical OS goes through the boot process and establishes a sort of state-machine ok i see your question and just by looking at the OS in its normal running mode, you couldn't reconstruct the boot process from the evidence see, this is why "OS" is such a bad term because OS is distinguished from bootloader what I am getting it is, would the reflectivity of tunes allow it to rewrite its own boot code if it proved useful? of course that's why we call it a system and not an OS it sounds like every PC running tunes could be running a radically different system any code in it (which includes bootloader code) is part of the system and "remembered" yes it would there have been systems roughly like it... the os on lisp machines is an example... all written in lisp, from bootloader on (however the HW was designed to make it very easy to do that, so they cheated) so, there should be some kind of critical mass. you write enough code by hand to define the tunes system, boot it, and it begins to reflect and modify itself (as a deterministic or statistical result of the intiial code) until it achieves a certain level of necessary usefulness well perhaps not that the process ever ends there's the possibility that your initial language has too much baggage to let go of via reflection starting with c obviously doesn't help or linux would be Tunes-- :) yes.. it might or might not be good enough. c is not that's why i'm proposing and designing slate well, at least it would be a hell of a nightmare. you'd have to write a system in C that could extricate itself from C which is an incredibly simple language... sort of a uK language of sorts i'm sure the XML code standard people will want to abuse it abuse what? tunes? slate oh they won't be able to handle it oh, good they'll foam at the mouth or something if they try to standardize it (seriously, it's that wierd) it's designed to be self-modifying, or at least support self-modifying systems? * water/#tunes nods it can do all the things that lisp can do, but also acts like a file system of sorts and it modularizes everything and has less built-in assumptions i.e. it allows you to throw out things you don't want or like pretty easily the semantics aren't fully debugged yet, though 03:00am the curious thing, to me, is suppose you/we/whoever succeeds in creating the intial state of tunes so that it can modify itself sufficiently.. who knows what it might surprise you with. by analogy, I think Benoit Mandelbrot was rather surprised by seeing the Mandelbrot set when he chose to graph z' = z^2 + z hm maybe so you could boot this thing and come back the next day and not even recognize what it's doing i know i've been surprised by what i can do with slate already well you can always meta-program it to stick to a particular way of doing things i've seen some state-machine programs that start with a boring 2d display and apply some rules for the stepwise evolution of the pixel colors, and they produce some totally unexpected patterns of course. hm yeah i know what you mean i mean, it's a good thing, since you know it's going to evolve towards greater usefulness (unless some nasty bug in your code makes it do something else entirely) tunes bugs might be really onery yeah. especially with the migration capabilities. suppose a rogue system somehow evolved a virus that was able to pass through a bug in the tunes security model? of course, other tunes systems might learn to defend themselves if they did not share the same bug onery? well Fare's always said his favorite tunes game would be to crack it :) yes, I saw that ;) an even better game would be to teach tunes to crack itself (or your neighbor's tunes PC) ornery: Mean-spirited, disagreeable, and contrary in disposition; cantankerous. See Synonyms at contrary. -------------------------------------------------------------------------------- oops s/on/orn/ got a little carried away with my mousing :) yeah i mis-spelled as well i noticed :) i thought onery was the tunes unification model heh can we put it in the glossary? heh ornery or the tunes unification model? i suppose youd have to define both hopefully the glossary and review upgrade will be finished soon nah, the glossary needs other terms much more than those i suppose we'll get to them, anyway oh, quesiton: it seemed like the links to your docs about arrow and slate were only .ps files. is that true? for arrow, yes for slate, nothing is in .ps yet... all html 03:10am and .lisp for the tiny evaluator i misremembered slightly then. will you make the arrow docs available as something other than .ps? yes, most likely the new papers will be in sgml and converted to both .ps and .thml s/.thml/.html/ great if i can get my web-coding friend to stop chugging pepto-bismol and start helping like he asked to should happen this weekend * water/#tunes crosses his fingers hmm. i'd offer to help but i don't know a thing about sgml he knows the ins and outs of a lot of tools but he's a lead sysadmin at an isp so he's usually busy is there a plan for getting tunes into production? gosh, I could almost imagine someone saying "leveraging TUNES" one of these days ;p hm well Fare and i seem to be converging on plans everyone else is basically waiting for us i think eihrul's working on the slate compiler (mobius) for example, the ideas about interoperable modular components, that can automate each other freely (and the GUI could just be seen as a way to expose the automation interface) freely scriptable yeah that can be done right now on windows, unix, etc. well, even squeak has a freely scriptable gui made of fine-grain objects it even has a new module system, though it's not robust just yet ok. in my work I'm forced to approach this from the practical side of still needing to run windows for the apps heh you don't need windows for squeak i've run squeak on my pda, of all things you see, what got me started on this whole thing was trying to run linux and windows on a dual boot system ah * water/#tunes has 3 oses he can boot into on one machine and i couldn't find a decent way to run a mail client under either OS that would use the same mailbox files interesting that squeak has a mail client in beta right now it pissed me off enough to want to create a universal cross platform user agent which i soon realized was almost the same thing as a universal cross platform OS you could definitely code up an agent in smalltalk and have it be cross-platform via squeak and all ways lead to the center... what about gnus? gnus sucks serious ass water: oh, i love it :) compared to squeak? 03:20am gnus? is this some uncommon plural form of gnu? no, gnu smalltalk he means scipient: gnus is the news/mail client in emacs oh!!! i did NOT mean gnu smalltalk no :) n/m then haha sorry for the confusion :) * smoke/#tunes crawls away well emacs apps are great if you like the emacs interface hmm you can run emacs on windows i guess but i detest relying on key-bindings to automate a stopgap measure at best, to my mind or not having variable-width fonts or anti-aliasing but then, windows and unix are all stopgap measures until people can figure out how to write a decent OS (or just S) can squeak run under windows or unix then? yes it runs on at least a dozen oses identically sorry smoke, didn't mean to put down your mail client earlier well, except where the os doesn't support certain features I presume a tunes system will still support programs written in C++, Java, Perl, etc., if the user insists on using them? heh yeah but we won't code it ourselves... water: fsf emacs 21 will have variable width fonts, and nt emacs has anti aliasing. still, it's an ugly mess. "if you want it, go code it" yeah, an ugly mess is right well, there will always be some practical value to the old-fashioned way of writing software heh * water/#tunes gives scipient the finger screw that i mean, isn't part of the reason that tunes doesn't exist already b/c it is much more difficult to conceive of? someday people will see today's style of programming as a disease of social control addiction :P well, you have some strong points there it's not difficult to conceive of, if you're not addicted to control water: if that was said about oil painting 300 years ago, i'd laugh the same i doubt oil painting holds the same psychological position as coding in c++ (photographing has taken over, but is by far as powerful as oil painting) but. suppose i'm using a mail client on tunes and i say to myself, hey, I want to be able to pull a line randomly from some [orthogonally persistent] textfile. since I am an experienced Perl programmer, throwing together a quick 3-4 line script to do this is just fine textfiles in tunes?!? some of the best painters made their own paint well, you'll have some way of storing textual information * water/#tunes slaps scipient around a bit with the tunes docs on word-processing 03:30am "since I am an experienced Perl programmer" hey, i just heard about tunes today you know see that statement i just quoted? yeah that's not what tunes is about tunes is about being able to code at the right abstraction level for the job i'm not implying that it should be about that. i'm just saying that some people will want to do that. not you of course ;) tunes will be The revenge on all perl hackers if i understand things correctly :) well i didn't mean perl specifically :) scipient: you'll be forced to at least use five parenthesis to perform your perl statement :) heh nooooo for some people, writing in Perl may be the perfect abstraction level for the task at hand no you're missing the point perl is not an abstraction level and frankly, I think all these symbols are rather pretty in the right order @{[ $_ ]} it's a dirty low level one that would tend to make tunes very brittle you could of course abstract the text to an ascii level, then modify it using perl, and then see what happens perl uses a certain abstraction levle slow down guys water: oh ok sorry "abstract the text to an ascii level" :P yeah let's coerce our data to fit the tool we already know!! :P why not? if that's what we want. tunes should be adaptable to our idiosyncracies fit the tool and the data to the thought yes it should be *able* to but doing so in general will make tunes brittle i agree that fitting the tool and data to the thought will be far more *interesting* and fun