IRC log started Tue May 16 00:00:01 2000 nah, people who work on the stuff we're working on get paid lots of money to work in well-equiped rooms full of hardware with connections to other experts sure, they also have PhDs :) not all of them [msg(TUNES)] permlog 2000.0516 * water/#tunes gives up on trying to sort out the 1:1000 s/n ratio for searches on modality sheesh, none of fred chow's papers give me anything more than a bib note off of research index no link to a home page? nope :/ jack pot! heh eihrul: does that mean "hcf: halt"? yes you're free hm hcf: hcf :) 12:10am * eihrul/#tunes counts the days until school ends. * water/#tunes counts the days until... well, you know :) eihrul: would this, http://granite.sru.edu/~chen/SSA/SSA.html be an ok intro to add to abi's ssa factoid? checking... hm all the gif's are gone yep, looks good damn she's up to 1058 factoids now with probably about 1% noise at most 12:20am -:- SignOff hcf: #TUNES (Leaving) -:- hcf [nef@207-172-225-84.s84.tnt1.pld.me.dialup.rcn.com] has joined #tunes -:- SignOff hcf: #TUNES (Leaving) -:- SignOff eihrul: #TUNES (sleep) -:- new [user3707@195.229.128.250] has joined #tunes hi new 01:10am -:- SignOff new: #TUNES (Leaving) boy the linux croud would throw fits about Squeak's GUI if they knew what it could do 01:20am -:- SignOff water: #TUNES (Ping timeout for water[tnt-10-136.tscnet.net]) -:- dirt [wrong@niantic0305.mohawk.net] has joined #tunes -:- dalvarez [dalvarez@tristan.sc.cs.tu-bs.de] has joined #tunes -:- SignOff dalvarez: #TUNES (Leaving) -:- thomas [thomas@193.217.63.152] has joined #tunes -:- water [water@tnt-10-193.tscnet.net] has joined #tunes re hey what's up? how's the vaca so far not bad i think it'll get better pretty soon, though 06:30am -:- SignOff water: #TUNES (The Tao went that-a-way!) -:- water [water@tnt-9-148.tscnet.net] has joined #tunes -:- SignOff dirt: #TUNES (Read error to dirt[niantic0305.mohawk.net]: EOF from client) -:- hcf [nef@207-172-225-206.s206.tnt1.pld.me.dialup.rcn.com] has joined #tunes ` 08:30am ' hey is there anyway you could elaborate on some hll stuff for me? * water/#tunes looks for the specific places he'd like elaboration upon blah the whole spec is so fscking useless as a starting point 09:00am why don't you make the whole thing an ANSI standard so you can be assured that no one will ever satisfy it? ok, down to business objects and attributes work context, however, is way too vague for me to get a handle on, unless of course you mean that it simply involves quotienting results * Fare/#Tunes is back i'm looking for a more specific take on your notions revolving around "context", "views", and "rewrite" my system can handle objects and attributes, both static and dynamic (this is all in reference to ./HLL/semantics.html for now) -:- SignOff smkl: #TUNES (brb) no, Fare is obviously not back what an ass well, i believe i have a framework idea that will work in implementing (at an abstract level) the tunes hll notion of context and what results from it 09:10am however, it won't be clear to use unless i roll a nice interface over it -:- smkl [sami@glubimox.yok.utu.fi] has joined #tunes re hello water getting useful info from Fare about tunes is about the most pointless kind of interaction with him Fare: if you manage to read this far, please elaborate on "choice" or give a url that inspired you to write that blurb on the semantics page about it hum oh, good, an actual hint of future response you mean the epsilon construct? it is to \exists as \lambda is to \forall ok so random assignment, possibly bounded ? what about the other things i mentioned? i definitely know how choice will work in slate, as long as i can get the meta-object system fully worked out you mentionned your difficulty with the notion of "context" no, with *your* notion of it ok it's too imprecise as stated 09:20am (btw add all this to the page, please) so every object comes with a declared semantics, that specifies the algebra of authorized rewrite rules on it i can support an abstraction which i *believe* is the same as your context idea, but i can't verify without clarification of your spec bs even though only a subset of this algebra would be effectively implemented so external behavior... how exactly should this relate to context? for isntance, a rational number is declared as something than can yield a couple of integers such that the ratio is conserved can you provide a non-trivial example? again bs too trivial but the underlying implementation will typically only remember one couple besides your ontology for that rational example is bogus wrt algebraic ideas you're not listening Fare *non*-trivial example, please the rational number idea can be done from too many perspectives to provide a clear example for context well, consider a game that runs on a given machine hence it's trivial in this context you want to migrate it the new machine uses different heuristics but that's ok as long as it satisfied the declared semantics for the game, too vague instead of emulating exactly the semantics of the previous instantiation no that doesn't work i.e. migration isn't about moving something with exact same bug-for-bug choice-for-choice compatible implementation you will have too throw out part of the declared semantics too often for that to be feasible it's about preserving the semantics thx for the sermon, Fare why throw out? it's precisely about the system not throwing it out unless specically asked to because of UI declarations or overall exprience, which *are* relevant to a game unless of course you refer to an abstract logical game in which case language still applies, and you will still have to throw out specifications when languages change if I migrate a quake game, I want to be able to change resolution and responsiveness this is too high-level, Fare you're talking about a huge example s/high-level/large-scale/ if you declare that some program use a RNG, and you migrate the program (you still have that asinine ping reply) you do not necessarily want it to use the same Pseudo-RNG algorithm but sometimes you do (unless of course, you explicitly declare so) 09:30am boy how much BS can you put out? in both case, the system must remember the declared semantics, all you're describing is a high-level language with a reflective implementation which is not as much as what's in your spec and you have no proof that such a scheme will work in the general case, because of expressivitiy limitations ! it might not work in "optimal" way fsck "optimal" but it will still work, if only because migrating bug-for-bug in an emulation box is still possible it won't work except in very limited cases unless you start thinking up new ideas bah this is useless you're selling snake oil on tunes.org I don't understand what you object. you're specifications on the semantics page communicate something more difficult than you imagine s/you're/your/ you need to re-write it in terms of effective programming language constructs I'll have to revise it, maybe ok pseudo-code examples would benefit greatly, at least lambda notation i think i've fulfilled your spec, then as long as i can get my meta-objects fully worked out (the whole library of them) great unfortunately, meta-objects as a field are hardly mature, so i'll be playing with some new and odd ideas btw how do you feel about my unifying scope with object structure? you know, the fact that my objects are inspected like contexts/namespaces 09:40am and can you like without an explicit lambda construct as a primitive (it's a user-level function in slate)? s/like/live/ * water/#tunes feels like he's talking to a wall hm i need to forward some language design decisions to eihrul's attention 09:50am to a curtain "curtain"? unifying scope and object structure is good well, except that scope frames usually also include information such as current continuation ok but how do you include the idea of remote access in the language? as long as lambda is _expressible_ it's ok don't care if it's not primitive ultimately you have to have a path to every object in the system, potentially (better be part of standard library, tho) right now, i'm using a directory metaphor, but this doesn't fit well hm well what do you suggest? also there's possibly a performance problem if the compiler can't optimize away objectwise access from usual scope access, if it can't merge scopes, etc yeah, lambda will definitely be part of the stdlib along with many other things ok you'd want to be able to "close" a scope, forbidding further modification to it well, i'm still concerned about the semantic structure of the language with this scoping mecahnism closing a scopt ... that's not a problem if your scopes are too open, and new definitions can be added at runtime, scope lookup becomes very expensive s/scopt/scope/ yes i know it's taken care of still... the issue of semantics and remote access of scope remote access? 10:00am do you have any suggestions as an alternative or improvement to my directory metaphor? not at the moment yes, right now it's done via cloning from remote scopes (read "cloning" as a null object as proxy) well if you consider an object system as like a directory structure... if you're within a certain object, and you want to access another object in the same scope that the current object is within, there are semantic issues in a unix shell, you would use "../the-other-object" but i'd like this scope access to fit into the language, so i make ".." and "/" part of an object's standard protocol but this is problematic, in that ".." seems to violate scoping directly do you see the problem?? (btw, "/" is not a problem since it merely places the evaluator in the object's context without evaluating it) i could make the scope structure totally graph-like, but then the orientability and understandibility of the scope structure becomes difficult to follow hum I think .. is evil it defeats many purposes of scoping such as security well, one alternative i've considered is to make the structure graph-like such as proper tail-recursion bah i can make it secure because the MO can intercept and void the message 10:10am what do you mean "graph-like"? Isn't it already? i mean not a tree for the most part if i make it graph-like, then there still must be a primitive for that closed-loop access somehow closed-loop? hm let me see if i can clarify why i think this is a problem there will be (eg) a scope with the natural numbers within it not instances of them, the actual things the only way you can use them is to access them from their creating scope s/creating/enclosing/ sure that's like opening a module well in order to clone them, you have to be able to get into that scope from anywhere in the system i'm not sure what "clone" means here 10:20am numbers have no identity (actually, you usually don't clone them, but that's irrelevant) they are pure values yes i know my cloning is self-style there might be standard repositories you create a new empty object that simply forwards all requests to the parent object except for meta-behavior in this case of course, but that doesn't affect the fundamental problem i mention there is no problem why not? the problem was to access the parent from the children there, you access the children from the parent yes accessing the children from the parent is ok yep that's what everyone does in every language yes i know this am I missing anything? i think so i think there will be cases where you will *have* to access the parent somehow from the children where why when? whether i formalize loop-like access (which is troublesome) or use a semantically-questionable slot are the two options i see right now hm an example, then let's say i have a running system where my natural numbers are in /system/math/naturals sure i'm working on something else in /toy-projects/number-games well, ./sieves perhaps ok so i'm sitting in the sieves scope, but i want some natural numbers how do i request it and how does the request get processed semantically? basically, what's the Right Way to handle this according to tunes ideas? hm either i've stumped him, or he's afk again 10:30am is anyone else listening? something like open /system/math/naturals well that's what self does or like /system/math/naturals/1 and there's a little problem with that ok, so what? you're allowing access to the root object from everywhere not exactly heh i'd like to hear the explanation, then you can explicitly rename the root and replace components in it, for objects below you explicitly rename it? how can you rename it if you can't access it within the current scope? on the other hand, if an object has already been compiled with some root, it will still use that root even if you rename your root i.e. the binding is static although if you have the right to decompile the object, you may try to instrument it to change its root ok you've lost me exactly: you can only change the root for objects _below_ your scope. no you're missing my point how do you get access to root in the first place? you have implicit access to it at compile time fuck compile time this is about semantics you can't escape the difference between source and object. grrr For if someone takes the same source and compiles on a different system, he might well obtain different results there are no "declarative" aspects to slate code because, e.g. the semantics of such standard function has changed it's emulated you're totally missing my point 10:40am I seem to be can you restate the point, please? can you show me where an existing language does it right or wrong? i'm talking about how does any object in the system get access to the top scope in a hierarchy in a way that fits into a tunes language hm your example is like that used in Self 4.0 * water/#tunes tries to think of a better example brb ok keep in mind that i'm using this scoping idea to help model distributed domains -:- Ghyll [karltk@msx-osl-15-54.ppp.cybercity.no] has joined #tunes in fact, i'd like to use it in just about any case where a set of objects is needed to be manipulated or inspected hi ghyll hi water. did yesterday's walk have any inspiring effect on you ? not really but i did relax water: that's even better. well, distributed domains are unlikely to have a "one" priviledged parent sure 10:50am and my scoping idea is very flexible due to its subjectness to meta-behavioral modification water: does that mean I can shove entries from one scope to its parent and vice versa ? well, we're working out the issues with that, but yes water: do you have a balloon ? huh? you mean a smalltalk balloon? :) I just want to mention 'name clashes', and hear it explode :) oh water: hmm. probably too deep a joke. that's not much of a problem water: how's that ? besides, what you mention is not primitive immutability is the default don't worry, though, slate can maintain many user-illusions good. btw, tonight's a beer-and-pizza session with a guy who wants to work on the tunes website and yes he groks the ideas and volunteered to help and is an experienced web-site designer wow. treat him well :) well we're friends already even so.. :) tunes.org isn't much to look at.. well we seem to share the same opinions about what's on the site yep :) 11:00am Ghyll: are you elric? heh no, you can talk to elric on irc.lightning.com Fare: is that an obscene adjective ? no, elric is the friend i was talking about water: ah. ok. Fare: no. I'm not elric :) (no, he doesn't deserve the name himself) hm he's afk right now, though hm that's irc.lightning.net , not .com lightning.net is an efnet server um i disagree 11:10am disagree what? n/m you were right 11:20am water: ur aware of http://www.liszt.com/chat/ right? yeah but i never use it 11:30am -:- ult [ult@user-38lcn4c.dialup.mindspring.com] has joined #Tunes -:- SignOff Ghyll: #TUNES (Read error to Ghyll[msx-osl-15-54.ppp.cybercity.no]: Connection reset by peer) -:- SignOff hcf: #TUNES (Ping timeout for hcf[207-172-225-206.s206.tnt1.pld.me.dialup.rcn.com]) -:- SignOff ult: #TUNES (ircII EPIC4pre2 -- Accept no limitations) -:- plonk [plonk@cpt-dial-196-30-178-47.mweb.co.za] has joined #tunes hello plonk may we do anything for you? -:- dirt [cogito@riga14.mohawk.net] has joined #tunes re dirt re re Hi water. Don't think so. Lurking and watching... :> ok another karmic theif :-] unfortunately most of the conversation today ended a little while ago indeed :) * plonk/#tunes *grimaces* anyway you can read about what we were discussing on the logs, if you wish it was more interesting than what's likely to be discussed for the next couple of hours *grin* And these logs... where are they to be found ? abi: logs logs are #tunes logs at http://www.tunes.org/files/irc * plonk/#tunes wihstles a merry tune * plonk/#tunes whistles too anyway, check out the website and the slate stuff, please water: why? did you just drop some acid? heh no, but i was planning on going out for a bit soon ahh enjoy :-] i'll be back at about 6pm PST with another fellow elric so we can discuss tunes web design i also have stuff to communicate with eihrul about slate 01:00pm -:- ult [ult@user-38lc6f7.dialup.mindspring.com] has joined #Tunes -:- SignOff ult: #TUNES (Leaving) * plonk/#tunes thinks you guys are a bunch of bloody nutters. Makes for a change. 01:10pm -:- ult [ult@user-38lc6bu.dialup.mindspring.com] has joined #Tunes * plonk/#tunes wanders off into the mailing list archives never to be seen again -:- plonk [plonk@cpt-dial-196-30-178-47.mweb.co.za] has left #tunes [] -:- hcf [nef@207-172-225-163.s163.tnt1.pld.me.dialup.rcn.com] has joined #tunes -:- SignOff ult: #TUNES (ircII EPIC4pre2 -- Accept no limitations) -:- mathias [math@france2.hemmet.chalmers.se] has joined #tunes -:- mathias [math@france2.hemmet.chalmers.se] has left #tunes [] -:- webHOGlogOFF [fdkg@adsl-63-203-156-101.dsl.lsan03.pacbell.net] has joined #tunes hello -:- webHOGlogOFF [fdkg@adsl-63-203-156-101.dsl.lsan03.pacbell.net] has left #tunes [] 02:20pm -:- eihrul [lee@usr5-ppp182.lvdi.net] has joined #tunes -:- Kyle_L [kyle@cr168790-a.nmkt1.on.wave.home.com] has joined #Tunes lo gakuk gakuk! 02:50pm -:- SignOff smoke: #TUNES (One day sheep will rule the world) -:- SignOff water: #TUNES (Ping timeout for water[tnt-9-148.tscnet.net]) -:- tcn [Tom@cci-209150250042.clarityconnect.net] has joined #tunes tcn! -:- SignOff tcn: #TUNES (Ping timeout for tcn[cci-209150250042.clarityconnect.net]) 03:40pm -:- tcn [Tom@cci-209150250053.clarityconnect.net] has joined #tunes ok, i'm really here now great what news of retro3? or is it retro4? what's stopping you? i'm thinking of working on it some more you know JonT@us.ibm.com? no someone working on L4 ? L4? never heard of it, not sure.. Jon Tidswell Advanced OS Technology Group / Sawmill Linux Project (thank you, google) apparently, he hacked GRUB yes, some of the L4 people worked there we got talking about a scheme for partly compile-time/runtime protection, which would fit nicely with the persistent object store I have in mind, and with forth.. mind if I run it by you? go ahead! 03:50pm ok, all the objects will be size 2^N, and pointers will be the segment+offset type the secure versions of the object-access words will simply AND the offset with (1< I hope there's a logical layer in-between that remembers the actual size of objects yeah, that's to be done at compile-time is the N variable? yes if so, how do you guess the "right" N ? bibop? umm bibop? bit-op? mmmbop BIg Bag of Pages well, if an object is allocated as 13 bytes, 16 will be allocated.. round up to the nearest power of two abi: bibop is BIg Bag of Pages, a memory allocation trick whereby you retrieve an object's size by a lookup on high-order bits of its address does it require paging hardware? tcn: what about "administrative bits" for an object? Are they to be out-of-band? no, bibop does not _require_ paging hardware (although paging hardware sure helps) I remember that old emacsen used kind of bibop by using the stripped high 8-bit of an object's address as a type tag (stripped by the hardware on a 68000, by the paging setup on other beasts; or by a manual AND on other machines) indeed, abuse of bibop can cause bloat in page tables I don't care for that idea.. (because of a sparse address space) 04:00pm tcn: that's ok i'd rather keep track of object name/size/location/flags in a dictionary or something http://www.xanalys.com/software_tools/mm/glossary/b.html#bibop abi: bibop is also at http://www.xanalys.com/software_tools/mm/glossary/b.html#bibop okay, Fare. tcn: you can end up with bloat in dictionary, too yeah, not dictionary, just a table of unnamed objects (a dictionary looks mostly like a reverse page table) -:- I440r [mark4@purplecoder.com] has joined #tunes hi yall :) lo so everyone here is a fort coder or what... anyway, the AND (1< bummer! we're just dreamers in here fort? tcn: am I dumb. (1< yeah tcn: of course, the AND doesn't _check_ for overflow, and you might end up clobbering stuff that shouldn't (which might be particularly bad in some cases= right, but it ensures that only the given object gets clobbered that may or may not be acceptable, depending on the rest of the system. you think I could get away with that? in some cases, yes not in all cases, tho I mean, it's a truely nice hack do checks before u do the (1< but I'm not sure it buys as much worth as it cost what values of n would cause errors ?