TUNES: No such nick/channel [msg(TUNES)] permlog 1999.0826 TUNES: No such nick/channel * Plundis/#tunes is gone. Gone since Tue Aug 3 18:38:00 1999 hey fare you awake? 08:50am Hey Anyone alive?? 10:30am hey niihau, sr 11:20am what's special with www.linux.org? nothing D: argh enough.. sr: dont change the topic w/o reason 11:40am * Fare/#Tunes is back >>> sr [depag@phila-dialup168.nni.com] requested FINGER from #tunes The party of the first part shall be known in this contract as the party of the first part. great idea... hmm (Groucho Marx) 12:20pm Hello :) hi 12:30pm I study TUNES web-pages for several weeks and I definitivly don't see how a TUNES session shall work : how the system should switch on ... how new features should be included out of a dk or net. Where can I find some run-time examples ? * binEng/#tunes passes the q on to fare ...or anyone :) I guess a lisp-like prompt would welcome users when login... am I right? Would end-user be authorized to write applications with the LLL? 12:40pm looks like nobody else is here... you're out of luck * Miellaby/#TUNES simply cannot be lucky. :) bingEng : perhaps you could try :) mielle: depens on configuration depends I expected the answer to be something like that :P probably there won't be any difference based on what lang the programs are written, if they are type checked, then they can be ran safely OK. Now, my main source of fears is the compatiblity between existing TUNES systems... how can we be sure than a personnal program will work on a friend's TUNES ? well you cannot really be sure, but the dependencies can be checked and the needed software can be installed (perhaps automatically) _ I read in Fare's pages that TUNES needn't any redirection layer when identifying software objects. Is that mean that as soon as a function name change, all the calling functions would not work any-more ? 12:50pm * Fare/#Tunes is back a dk? a net? My question is : When importing a new software, are the references to the existing software material resolved only once time or each time code is executed ? Miellaby: seen the oberon interface? Fare: You think This could help me ? I didn't Fare: I can't manage to imagine a system where you can't 'physically' handle a same function whith two different aliases. Miellaby: difficult question. it's actually the single most difficult question in dynamic distributed systems, or in life: what's the identity of an object? what are its limits? and there is no canonical answer. systems that try to enforce a canonical answer (as in "always early", or "always late", are doomed. 01:00pm toujours la? Why not using systematicaly an intermediate id. layer ? like UNIX inodes? to make sure that a program works as desired, you must basically define yourself the limits of your objects. metaprograms provided with the system should help you draw the limits dynamically but ultimately, the decision is yours. I fear metaprograms are not the ultimate solution. as always, they are not the end of problems, only the beginning to solutions. (I use this sentence for free market vs economical problems, free software vs computer problems, etc) I d'just like to put the stress on the fact that even on low-resolution system like Windows(c) there can be more than 10,000 files/objects. Howmany entities will co-exist on a TUNES session? hundreds of thousands; millions. most of them passive, happily. also, it will depend on your point of view. I'd like TUNES to not be a single-minded system; To my mind, TUNES objects should be park on a tree graph. then.... any "object" supposes that you have chosen a particular point of view under a partial, high-level, point of view, you'd only see very few objects under a global, low-level, point of view, you'd see a lot more objects, but simpler ones. I think current systems are fundamentally wrong in trying to having a "single level" for everything they don't manage to do so, and end up having several levels, in a klugy way (shells vs applications under Unix) 01:10pm So, I'm a TUNES newcomer. That's the first time I'm connecting to a TUNES system. I want to "walk" and discover existing TUNES entities. How could I do that ? you'd start from the initial objects in your current UI, and discover the system from it for instance, you'd select an object, and query the system about applicable methods to apply on it (typically, by clicking on it) then you'd apply the method sometimes, you'll need push several objects onto stack to find an interesting method hopefully, you can browse the whole system from your initial objects (if not, it means you're a restricted user) You said there should be several waies of handling a TUNES system. I though folders/directories/groups... were THE tools to do so. when you can equally browse arbitrary structures, you don't need formal "folders" and "directories" you want to create a bag of objects? create a bag of objects; they needn't even have names! or if you WANT names, you can have ones. when you browse a bag on a terminal, it appears as a list; on a window, it may appear as random icons in a 2D windows; or as random objects in a 3D area (in a holographic interface) or random pitches in sound-based interface I'm perfectly understanding ... and how you could define such a bag? (you'd hear, e.g. well distinguishable sounds at the same time, and could recall one by imitating it) Miellaby: a bag is rather a trivial data structure can you give me some lisp-like line-command to do so ? the question is rather: how do you _tie_ a representation of a bag to an actual concept of a bag. and the difficulty comes from _intentionality_ 01:20pm Miellaby: I believe the interface would be rather point&click and reverse-polish, for every-day operations; you'd use lisp-ish syntax as soon as you want to say more structured things at once. (poplog too has the two kinds of syntaxes) point&click is generally much slower that hand-typed operations (at least in Unix) Miellaby: sure. I'd use RPN on a terminal and escape into LISP for more elaborate expressions e.g. 1 1 + and then :(some (lisp) (program)) i.e. some character (e.g. colon) would trigger lisp mode. since lisp is self-delimited, you'd go back to RPN immediately after I think that stack model is a old-fashionned way of handling computing ;) stacks are useless ! (and of course, you'd have completion in either mode) Miellaby: I never found a better interface than the HP 28! add powerful line editing to that, and you're done! but of course, you'd be free to make your own shell, if mine doesn't suit you! 01:30pm not that you can't do that today... I don't want to write my own shell. Even at the lowest level, stacks such as Context Stack just should not exist. They should be substituted by a set of temporary cloned frames. The ultimate OS primitives are, from my point of view, com-like st ructure managements linked by referencing and refining relationships. :)= comb-like ... comb? binEng: what can you do? binEng: you sure can't do that in Unix... make a new shell? bE: sure, but the shell you make is limited by the underlying architecture I wanted to write a natural-language shell for Unix some times ago. the point of tunes is not so much in any particular shell than in the architecture that that shell interfaces to Miellaby: there is nothing to interface to in unix but a a filesystem that contains lots of files among which a bunch of standalone programs that must essentially communicate through those files Fare: I imagined some musical examples of what my own idea of an OS should allow to its user:) Miel: there would not be a One Context Stack, but every shell "window" would have its own context stack Fare: sure, I was just making a point :) bE: thanks Miel: par telephone, c'est plus efficace... oh! je n'oserais pas! moi si, vu que j'ai un forfait Fare: I just don't want. I'm shy. just try! what've you to lose? by staying on irc, it's time. by trying phone, it's ??? M: if you fear that I'll call you night and day until you lay in bed with me, you can calm down -- it's not my style What a pity ! 01:40pm I read that TUNES channel is continuously traced. I want my ideas about world conquest to be written for the future generations :) so sorry ... 01:50pm I remind this story of bag... I suppose I can implement it via a single lisp list? I suppose that each atom in the list would point to an other defined entity. there is a paper about programming with multisets in escher in escher? multisets? * Fare/#Tunes is back abi: escher? i think escher is a functional logic programming language at http://www.cs.bris.ac.uk/Research/Declarative/escher.html traced? Fare:recorded. http://discus.anu.edu.au:80/~jwl/ Miellaby: if you want what you say to stay, put it on a web page (possibly the tunes web page) nobody reads the irc log unless specifically searching from something precise i.e. it can be used against you, but don't expect people to read through all of it until they find your quotations. Fare: alta index'd em hcf: ouch! 02:00pm Since I found TUNES pages for the first time, I have read my own essais about my on projetc. My product looked so ungenuous in comparison with TUNES ! I'll never write a paper again; im ~wrong, yahoo index'd em smkl: I guess his next programming language will be "Bach" well, they'll need to implement escher first miellaby: don't say that -- instead, improve! miellaby: when I re-read my first attempt at defining a system, I laugh! You should put it on TUNES site :) my, I wanted a cluster of 6502 with an MMU (I didn't know the name MMU, but I have reinvented part of the concept) (I did know, of course, about the way the Apple ][ paged memory about 48K) s/about/above/ 6502... yum ! A good OS should be able to work on such a small ship. depends on what you call "to run". "to creep":) I'd rather say "to take advantage" -- i.e. a main host runs on the main computer, and lets the lesser chips do part of the work. although in the case of a 6502, managing it might cost more than it would pay back! why so much people are mad of distributing software around computers. Is this really useful? Miellaby: I work at FT. It's not a question, it's an axiom. CNET. I see. Our system IS distributed, there's no question about that. * Miellaby/#TUNES worked at CNET in Lannion. 02:10pm oh? Issy and you're afraid of phone? maybe that's why! I'm only a humble coder. no coder is humble! Miellaby - The Only True Humble Coder Shy and humble. Yes I am. ;)~ * binEng/#tunes warships the One and Only ;) \_ (speaking about something else) The basic building blocks of an OS system should not be abstract symbolic lists but concret small structured memory areas, cut into parts by local ids. some of these parts should be pointers to others memory-c ells. other parts would be raw material. And the header of each of this ... these cells will be a list of some mother-cells which own delegated I/Fs. 02:20pm Tips wanted. Is there any way to do recurse-edit with IRC tools? I wish I could stop my sentence, quickly answer to a joke, then continue what I started. warship? let all battleships self-destroy! Fare: joke hey, that's an idea: convince the navy to put Win2000 on all battleships! (hum, maybe not on the subsystem that manages nuclear weapons) y0h well, I heard they had NT on a aircraft carrier... NT crashed and so did the carrier? it was left without propulsion for some time did they understand that no one in one's right mind should use NT? * Miellaby/#TUNES wondering : Should It better look at Oberon or Oberon2 ? Fare: it didn't say ;) 02:30pm What will hapens when the whole dictionnary will be used to identify those thousand of millions of TUNES entities. I think there's be a problem. Instead, TUNES should have a reference scoping mechanism. Like '(computer.driver.screen.print ...) Miellaby: speaking of yourself as "It"? reminds me of a Stephen King novel... scary! I'm a kangaroo :/ That's why. But it's an other topic. * Fare/#Tunes calls hopital Sainte Anne... Miellaby = Miel + Wallaby :) arf. This time, I'm really sorry. Miellaby == mieline abimee! why _memory area_? the user should be abstracted away from that! why a dictionary? why identify? identity is a precious resource that shouldn't be wasted. it's an ingredient that doesn't fit in parts of the system, either or course, we should have scoping, however. 02:40pm mieline abimee : It's helping me having some imagination / I think the best way using a computer is to accept that computing concepts are not virtual (as they are used to say) but absolutly concret. In a computer, even the most abstract concept s have to be put in some memory delimited areas. I think that neophyte users perfectly understand the concept of "concret objects". The logical layers to hide this simple rule are just confusing human http://www.ae.utexas.edu/~rwmcm/quotes.html A lot of classical computing concepts are very confusing : multi-processus, multi-windows, multi-thread. And now, distributed systems ! in a brain, even the most abstract concepts have to be put in some nervous delimited areas. I think a GUI (for example) should always apply the LAVOISIER law. notice that computers are linked to a power source precisely to avoid that law. ... and let the user command the "multiplication" of objects. I admit This's a 'fundamentalistic' point of view. 02:50pm * Tril/#TUNES reifies the meta-tower oh crap! what have I done * Tril/#TUNES ouches as the meta-tower kicks his ass! * Miellaby/#TUNES realizes what it saids has not a lot of meaning. :( Tril: logger not making new files, appending to 0824 * JOIN " TUNES #TUNES * " umode -s NORMAL 0 * MSG " TUNES NEWLOG * " set logfile $2 SILENT 0 * MSG " TUNES PERMLOG * " exec chmod +r $2 SILENT 0 * PUBLIC " * #TUNES * " @activity=1 QUIET 0 * TIMER " %:?0 " baz SILENT 0 * TIMER " *00:00* " exec -msg TUNES date +"newlog %Y.%m%d" SILENT 0 * TIMER " *00:01* " exec -msg TUNES date +"permlog %Y.%m%d" SILENT 0 >>> You(TUNES_) are now known as TUNES now it's fixed. it must be named "TUNES" or it won't work using the time stamps, someone can split the log into days again.. i don't feel like doing it Tril: or just rename the file 1999.0824-0826 better yet, make symlinks from each 1999.082[56] to 1999.0824 and be done with it 03:00pm Tril! * Fare/#Tunes meta-ifies the towel ...are you reading ALTTLG? I'm in the introduction Fare: hows the HLL? hcf: not hcf: I'm debugging my scwm configuration s/scwm// Fare: i guess thats vastly more important abi: by hcf abi:eby hcf abi: be hcf abi: ebay tril good job! abi: you are right, hcf is usually idle? tril: i don't know Tril: and? ? Tril: what was the point of that? seeing if you programmed more "Be" commands? do some bots automatically learn those? they needed be "programmed" by observing the channel? just factoids abi: be nick is something abi: be nick something i know that abi cant do it but I heard somewhere that does abi: forget be nick hcf: I forgot be nick oops, killed the wrong scwm! Tril: how do things get chosen? s/needed/need not/ 03:10pm hcf: I do not remember. Nor where I got the idea from i just remember somewhere, i think on a homepage of some IRC channel, a list of channel regulars and some excerpts of a bot imitating them Tril: perhaps efnet/#perl I though about a problem with reflexive OS. How can we make deeped changes into the system whithout going across some unpredictable states? what kind of unpredictable states? will TUNES provide some transactions primitives : Lock, Commit, Rollback ...? For instance : You want to change the I/F of a rather fundamental functions, used by a very big set of calling functions... Or better : you want to change a basic type of information : like the unit of a shared variable, a quotient, like the screen resolution... During a short period, you have to lock all the system while making the I/F changes and updating all the calling functions. Perhaps not the whole system but a subpart of it. or you can emulate the new system while the old one still runs. Have everything happen simultaneously in both, then when you're ready, replace the old with the new in place and no one will notice. That's an idea. Like swapping screens when drawing. m: at least for locking, all dependencies will be tracked so the proper areas can be locked I persit to say that all could be much easier if TUNES was OO at the lowest level. what kind of OO? how do you define OO, from the many ways? 03:20pm A way to gather dependent informations, in order to simplify updating operations. A way so that functions -at least unary functions- have to be glues to the type of entities they apply to. sounds like what I have been doing (I guess that functions wich involve several kind of entities should be cited in each concerned entity definition) What have you been oing? I am designing a type system for a persistent store, which has automatic dependency tracking ggg. ambitious ... it has to be, because it is meant as the framework for tunes wow .... I tried to convice Fare that TUNES should implement in its lowest fundations the operators to build a data structure which could implement a graph of object (whith refining relationship). I missed. 03:30pm i think everyone agrees to use a persistent hyperlinking store which is made of only references to within itself (me and water, anyway) count me in Fare likes persistent hyperlinking store? ok except that you should replace "only" by "mostly" hehe what exceptions can you think of? and I'm not sure what hyperlinking is; looks like an interfacing issue rather than a persistence issue. Tril: communicating with the external world!!!! that can be done by making hooks that appear as if they are "inside" the persistent store. how else would you do it no, that cannot at one point, it must be made user-controlable the fact that the external world has not the same safety properties as the internal world by definition, you control what's in, and not what's out. it's just like UNIX communicates with the outside world using sockets which appear as files no, it's deeper than that it's about the invariants you can have or not UNIX basically provides no usable invariants, so it's up to you to forge yours, and you see no difference between local and remote stuff -- everything is unsafe well, i object to external references- they cannot be guaranteed referential integrity TUNES does provide invariants on its managed objects, and thus makes a difference between local and remote -- that's the whole point of reflection 03:40pm of course, you can try to merge domain, and "fantasize" about it, giving results that are conditionalized by the non-failure of the fantasy. you NEED external references. but you NEED be able to conditionalize them. you NEED be able to fantasize. you NEED be able to make the difference between fantasy and reality that's just the difference between levels of interpretation (emulation) anyway, i'm not sure we're talking about the same thing 03:50pm but I know the problem I'm talking about (it's precisely what my boss wants me to clear up as my PhD thesis, incidentally :) I'm not sure I know the solution, tho look at ALT.* ... each administrator chooses his/her own policy. you may carry no alt groups (inconvenience). Or accept newgroups and ignore rmgroups (big alt), or honor all control messages (chaos) you do the same thing with conditions about external networks (or whatever you were saying?)- depends on which host you are running. [you didnt answer my question, have you begun reading a lisp through the looking glass?] 04:00pm I begun reading it long ago but tomorrow I'll print it at work back in '91? no, a few months ago so why did paul post its url instead of you? ALT.* groups however, are read-only; they do not otherwise interfere with the system. Think about what happens with external objects that DO actively interfere with the system... I think the URL was posted on the list long ago, not? I don't remember (the URL has moved since) hehe.. i did a search of the MOOSE list for the word "reflection" and got 0 hits i cant find any old mention of this paper 04:10pm hum. Maybe I forgot to warn the list. I do remember contacting the author, because his site was buggy. TRIL * Tril/#TUNES hides * dtm/#tunes waves to everyone dtm is back after months of silence * dtm/#tunes whispers "i see you" to tril after months of shame of being unavailable to my good friends shame for whom? at the tunes projects I thought you ran off because I offended you for having left my email address dtm@hex.net still subscribed silently to the tunes mailing list tril i would never do that! i changed it to dtm@tunes.org, you may want to check that mail box tunes is one of the coolest things i have ever thought of in my life 04:20pm and i owe it to you guys, you wizards of programming and logic (well, on computers anyway -- not necessarily in conversation! ;) you know what i did last night dtm: Last thing I remember is a conversation where we got upset because you were not a "programmer" and therefore someone had the impression you could not "participate in the project" ... ??? i dont recall anything of that particular context, tril OK forget it, then afaik my involvement was quite appreciated :) i did prefix my comments with that assumption, that i would possibly not make sense to some, or not be entirely valid in my comments obviously. it's taken me days to try to write a "simple" shell script here at my job, that gets a document and does some error-detecting and correcting logic and it doesnt work :} i'll do things like write a simple 'for' loop that overwrites the main variable inside it, and be stuck on that part for 2 das days not realizing that i need 3 variables instead of 2 does tunes.org have sshd installed, or still SSL-telnet ssh1 ok so i guess i'll be considered a munitions trafficker then huh i dont have ssltelnet anymore, no one has needed it is tunes.org in the u.s. or still france only if you download it from outside the US? hcf: hi hoy, dtm hmm... main server is in US, we have a web mirror in UK abi hi bonjour, dtm hoy dtm tril do you have a free colocation / virtual host? yes Tril: who's provided it a generous ISP, which I now work at, nas.com i see are you in WA y btw i'm currently in a contract at Cygnus Solutions in Sunnyvale, CA i moved to silicon valley from TX Cygnus the makers of Cygwin?? indeed the very same I thought they were getting bought..? maintainers of gcc, gdb, etc THEYR'E HERE Cygnus makers of _GCC_ Cygwin is unimportant; GCC is important. right on fare Fare: I thought RMS wrote gcc Windows is a dead platform, anyway. hey fare :) good old Tom Christiansen wants to port gnu fileutils to perl! this would further minimize the use of cygwin Fare: heheheh good call Tril: he did, originally, but Cygnus are those who did a tremendous amount to make it the compiler it is. Tril: yeah like every other thing rms started, it basically hit the fan Cygnus made egcs, afaik, and only recently became maintainers of gcc (when the egcs fork merged back with gcc) integrating it into tens of platforms, and building the binutils, etc no, they did maintain GCC long before egcs Tril: dont know about that but you can check http://egcs.cygnus.com Fare: get my email? Tril: did you know that i originally moved out here to work at VA Research? why don't you? i worked there for ~6 months well, "maintain", not. They did *port* GCC to many platforms and make many enhancements to it long before egcs yes cygnus's main thing is porting any software to any platform so to speak 04:30pm * liar/#tunes thinks fare is ignoring him :) Fare IS ignoring you, but not on purpose... but because he is doing so many things at once no i think he is ignoring me on purpose liar: historically, Fare has been terrible at reading e-mail..just send it again musta been my email to him last night he is ignoring me because of the email i sent liar: oh, well, what is it about? i sent him a picture of himself with a beard :) liar: uh? thats 2 pictures so far Tril: what u want em all?