IRC log started Wed Jul 14 00:00:01 1999 [msg(TUNES)] permlog 1999.0714 -:- XeF4 [xef4@wbrady.wasilla.ptialaska.net] has joined #tunes -:- SignOff XeF4: #TUNES (bbl) -:- hcf [nef@me-portland-us745.javanet.com] has joined #tunes -:- abo [d96-abo@brown01.nada.kth.se] has joined #tunes -:- norman [user2304@mail.4js.ste.net] has joined #tunes -:- norman [user2304@mail.4js.ste.net] has left #tunes [] -:- SignOff hcf: #TUNES (Ping timeout for hcf[me-portland-us745.javanet.com]) -:- NetSplit: lackey.openprojects.net split from sterling.openprojects.net [07:20am] -:- BitchX+Deb1an: Press Ctrl-F to see who left Ctrl-E to change to [lackey.openprojects.net] -:- Netjoined: lackey.openprojects.net sterling.openprojects.net -:- Mr_Wrong [seanl@adsl-63-192-213-146.dsl.snfc21.pacbell.net] has joined #Tunes -:- eStormy [stormy@rain.futuresouth.com] has joined #Tunes -:- s_r [sr@phila-dialup486.nni.com] has joined #tunes i gotta put windowing back in my script...i'm tired of having 3 ir windows open 08:40am -:- SignOff Mr_Wrong: #TUNES (Ping timeout for Mr_Wrong[adsl-63-192-213-146.dsl.snfc21.pacbell.net]) stormy: mirc script? bx hmm Deb1an? yes hmm...i wonder if i still have the windowing part of my script...i took it out a long time ago why not just program a client from scratch? too much typing i've wanted to before i have a lot to learn before i can but i want to i'd use python people on #BeOS are talking about be source code and how some was pirated :) like the folks on #beos would know that would be awesome... to get beos source code 09:00am petros? petros? hmph...i don't think i have any archives going back that far something they talk about in #BeOS stormy can you get me beos source code :P i can't even get me beos binaries heh it pisses me off... i know people have pirated the source tree i'll never get it though found a 1997 archive, but i had already ripped out the windowing or hadn't yet added it? * eStormy/#Tunes growls. Stormy why not just get someone else's script and rip it from there? cause the other scripts suck (c: 09:10am Stormy are you writing an OS? i'm dreaming up an os stormos what are your ideas for an OS? earth operating environment i have 2800 lines of random thoughts in 2 dozen text files, no unified vision yet. wow i wanna come up with something completely new and fresh, but that's kinda hard. my brain is already stuck thinking in certain patterns i feel the same way your name is unique, you should live up to its unique nature :P would your OS be multiuser? yes reflective? multi-user, multi-tasking, multi-processing, secure, orthag persist, distributed, etc. etc. i have been thinking of dynrec in an OS have you thought of dynrec? rec = recompiling? yes if so, yes the OS modifies itself in runtime i had this much about it written: files need to automatically compile themselves when executed i dunno about the new-kernel idea mentioned in here yesterday what idea mentioned here yesterday? about booting new kernels without rebooting nice in concept, i just dunno how it would work. well you would separate the kernel into "lock chambers" where it would be isolated so it doesn't overwrite what it is executing or "frames" 09:20am estormy do you want the OS to be truly object-oriented? * eStormy/#Tunes feels tao-flashbacks coming on... ? tao? well, tao is an expression of the Water in comp.os.misc do you practice tao? there is a 401-article thread on an os called tao what is Taos? 400 of the posts are flames because the guy wanted his os to be object-oriented, but he failed to mention his definition oof object-oriented. (c; tao is at http://www8.pair.com/mnajtiv/tao.html estormy do you see any OSes on the front that have much promise? beos is the most visible one. i don't know enough about the current stages of anything olse i want something as good as BeOS but opensource -:- AlonzoTG [Alonzo@209-122-253-198.s198.tnt8.lnh.md.dialup.rcn.com] has joined #tunes -:- smkl [sami@CMIV.rdyn.saunalahti.fi] has joined #tunes beos has different priorities (multimedia), but i don't really see anything new about it. (i haven't used it) beos is nothing revolutionary just really fast i guess you could sum up my basic goals: a new graphicial unix (retaining a powerful command-line) with no resemblance to unix. lotsa small tools to be combined 09:30am afk smkl have you seen the TaOS posts on comp.os.misc? -:- SignOff eStormy: #TUNES (Ping timeout for eStormy[rain.futuresouth.com]) 09:40am hmm .. i don't read comp.os.misc wait a bit -:- eStormy [stormy@rain.futuresouth.com] has joined #Tunes stormy: if i execute a "rep" instruction (like rep movsb or whatever), do interrupts still interrupt the instruction? i don't knem asm know perhaps you can use cli 09:50am TaoS sounds a lot like TUNEs 10:00am * Fare/#Tunes is back fare how many cycles does a "rep movsb" take? depends on the CPU if i had an interrupt, say clock interrupt rep movsb is generally a bad idea but on the 8088, tho execute a schedule() would the interrupt interrupt a rep mosvb or let it continues contine continue s_r: the Pentiums and others have an internal clock whose value you can time and then interrupt after that instruction is finished? see Michael Abrash's books for more info and/or tutorials on the web start from the pointers at the end of the Asm-HOWTO (latest edition) mind that caches, TLBs, etc, will affect the results greatly... (and thus of course data alignment with word/cache-line/page/pagetable/etc) 10:20am it is said that on plain pentiums (not earlier or latter designs), it is best using the FPU for moving large chunks of memory will interrupts interrupt instructions right in the middle of their execution or will it wait for them to finish? you mean rep movsb? any instruction... well, since rep movsX is a tail-recursive instruction, it may be interrupted in the middle, by saving the current registers, and pointing to it as next instruction ok it is tail-recursive... for other instructions, the hardware must either roll back or roll forward, so that you are always between two instructions when a trap/interrupt occurs -:- binEng [Anders@dialup211-1-53.swipnet.se] has joined #tunes if it weren't tail recursive, that'd be a problem (actually, you might have had short atomic reversible prefix/postfix to the tail-recursive part) 10:30am -:- tmf [s720@pissemaur.ii.uib.no] has joined #tunes tmf! hi scandinavian night, tonight! how's that? i wonder why everytime i have an mp3 playing, whenever i use a scroll-down bar, the mp3 is warped for the time i click the scroll bar down (in windows 95) is it because the scroll bar moving rapidly down needs a rep instructino instruction ? no seems like you have a slow computer, eh? it's because win95 thinks the scroll bar is a high-priority task ok (maybe because it runs inside the win95 kernel) s_r: it's a effect, a feature - be greatefull! bE: even on a fast computer, the scroll bar will eat loads of CPU yep - great stuff, this windows hmm that's a good way to look at that... a feature! Fare: not on computers as fast as mine ;) PIII-600 ? Fare: Alpha 700 um win95 doesn't work on Alpha 700, except perhaps in Bochs windows NT does though well not exactly. But I can type very fast on 'em ;) and it's not fast! i was saying that bineng might have an alpha fare: is there any way to make the scroll bar movement eat less CPU? just use a rep instruction to copy pixels? maybe with a quad PIII-600 and a RAID cluster of U3W-SCSI, Win95 might have barely acceptable speed but there isn't much difference in how X handles scroll bars and how win95 handles scrollbars isn't scrolling done by the video card? s_r: again, rep movsX is a bad idea. Use library instructions for bcopy() and suches, that will hopefully be optimized for your CPU s_r: yes, there is a big difference -- scheduling i wonder how bcopy() does it bE: who knows, with win95!? :) bE: bill gates could tell you, but then, he'd have to kill you maybe all output is being routed thru Redmond? :D fare: how does bcopy() copy it? does it use rep *? 10:40am s_r: just decompile it or read the sources note: the linux kernel has its own versions of it, that may be better than the versions in the libc... * s_r/#tunes looks at VSTa bcopy.S Will the fact that a system is _open_ jeapordize security? ( sorry, slight topic distraction :) .align 4, 0x90 movl 0x10(%esp),%edi what does the .align statement mean? what does 0x10(%esp) mean? (esp+0x10)?? align: locate next instruction at address divisable by N s_r: RTFM! i knew you'd say that WATFM! M-x info g as read the assembly HOWTO! howto? err where * Fare/#Tunes hits s_r with a truckload of intel manuals ahh intel * Fare/#Tunes hits s_r with a HOWTO compendium I think there is a point in trying and failing a bit on your own before understanding other peoples assembler i have one of those s_r: get a more recent version of the asm howto than printed HOWTOs! hey, where's abi!? abi: abi? hmmm... i am the little girl borg with a big brain or #tunes' personal little whore or gaklosmontic or flurivostuginuous or xyvarestoplik or an Artificial Idiot or braindead or a crack addict infoslut or stubborn abi: HOWTOs? HOWTOs are at http://metalab.unc.edu/LDP/HOWTO/ abi is an infoslut???? 10:50am -:- s_r [sr@phila-dialup486.nni.com] has left #tunes [] I'm reading the document on Tao. Tao? hmmm... Tao is an expression of the Water tmf: URL? IS IT possible to have an open AND secure system? http://www8.pair.com/mnajtiv/tao.html I mean, isn't "security" in itself the restriction of access? tmf: sure security is not paranoia to be sure of something, you needn't close your eyes and cover your ears to prevent someone from convincing you that the thing isn't true! no, but you eventually end up having to "protect" certain things from users or programs sure. So there will be programs that "would like" to be able to alter or view things - what is the natrual thing to do when a request for altering something is refused due to security? MU tmf: search for W7 on Review/VMs.html ok, I will. 11:00am The thing is that in TOOL, the natrual thing ( least coding solution ) would be to just give the object that modifies its own personal object, so that it can keep on reflecting (:)) without violating security. I believe Abadi has written a good paper in a recent book on the topic read the W7 paper and be enlightened! tmf: beware!! haven't you followed the thread on usenet over tao? no I'm just reading about it now tmf: the author is little more than a moron and the paper is pretty much useless, apart from acting as an example of how not to do/think don't waste your time on it ok :) so I should take more notice of the disection of Tao than tao itself :) That's true for many things the usenet discussion opened up my eyes. that text contains no more useful info that "viruses should not be possible to make" I noticed that too. How would viruses be impossible to make? bE: did you read the W7 paper? tmf: go read it, NOW! w7? no, what's that? -:- Kenn [cpe2@1Cust55.tnt3.tampa.fl.da.uu.net] has joined #Tunes the semantics of a language provide functionality the syntax provides usability ( The thing is that I'm thinking through the mapping of TOOL to asm and the comunication of asm with TOOL and vice versa, and these nice things keep comming up :) TOOl? Kenn: hi! hello Fare TOOL is TMF's Object system on top of TML what is TML? i think TML is a rewriting thingie, and TOOL is just an extention of that ( terms like 'object' defined ) rewriting thingie? abi: no, TML is string-rewrite system by TMF, who has implemented TOOL on top of it fare: no idea * eStormy/#Tunes fires up twin mp3 encoders on a hard drive of .wavs. TOOL is a useless inevitable extensiOn Of computing Linguistics:) 11:10am computing linguistics? I needed an 'L', ok ok tmf: have you seen Maude or CafeOBJ3? what is the nature of the obsession with recursive acronyms? tmf: you absolutely *must* see Maude and learn it is it a psychosis? what's good about it? K: it's a recognition sign between hackers K: and a hommage to recursion, the mother of computer science OK: Okay is oKay tmf: Maude is what TML should have been so Fare, how is progress on your scheme evaluator? And what is that? tmf: rewrite system of graphs, rather than of strings Kenn: not much 4 days left to finish an article for popl Maude sounds like Brian's Arrow system is it similar? Kenn: not AT ALL what is the utility of using graphs? Maude? Maude is at http://maude.csl.sri.com/ or based on re-write logic. sorta like TML, but much simpler to understand Maude is also based on well-understood, well-formalized theory okay, Fare. I sure as hell didn't think so! :) Maude is also reflective okay, Fare. is Maude intelligent too? FOO! Fare, I have a question for you. If syntax is only important in the context in which it is used then shouldn't the semantics of a language be defined so as to be contextually extensible, providing not any particular semantics, but a meta-semantics? Maude has alot of operators and syntax and stuff, that are hardly orthogonal and extensible. I am not talking about Maude Sure, but it has the problem that you point to. what problem? hmmm... problem is you don't have a clue! who? Kenn: yes and no Kenn: there should be a framework to define extended semantics but it should also be possible to make it clear what semantics you're using for a piece of code why not just make the language the framework and just provide a default semantics exactly, by specification 11:20am Kenn: the HLL should precisely be that think about persistence: if you want orthogonal persistence, somewhere something must still explicitly manage objects How do you go about defining semantics prior to semantics? instead of defining semantics specifically, define it generally There are no semantics to define anything from. generally or specifically but there are things that all semantics have in common like contexts of evaluation YES - BUT WHAT ???? every language evaluates expressions in a context Kenn: I'm not sure what problem you want to pinpoint EXACTLY !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! the problem is I cannot do low-level programming in typical high-level languages Kenn, add 'terms' to your model and volia - you have a meta translator ! what are terms? word, like '+' or 'object' or 'create a new term' for example term as in terminology Kenn: Tunes needn't be typical but how do i escape my model? hcf: around? Since you now have a meta translator, you can "send messages" to other contexts, like assembler, or an interpreter. how important is persistence? is it fundamental to the HLL semantics? Kenn: most important why? Kenn: because it means that you can make persistent assumptions Kenn: did you read my paper on metaprogramming? Kenn: if you deepen the complexity argument, Fare: I read your paper on non-determinism, reflection and the lambda calculus I'm just thinking about what the HLL should do you see that a certain class of "universal" languages (including FORTH and LISP but not C) are maximally efficient in terms of incremental development how can i alter the evaluation of expressions in LISP without writing an evaluator? but even then, you get a horrible computation-time behavior if the system is not persistent I understand why persistence is useful, but how is it fundamental? Persistence is what allows the system to not recompute everything from scratch as the system evolves. but what about the contexts in which persistence is unwanted? which transforms a linear incremental cost into a constant incremental cost 11:30am Kenn: if you're superuser, you have access at multiple "levels" of your system. Kenn: and can access persistenc infrastructure manually if you like Kenn, an object never outlives its context. why can't i do it without being superuser? tmf: exactly Kenn: LISP provides features so that you don't need rewrite all of an evaluator from scratch to alter expressions tmf: garbage collection is meant for that Kenn: ever heard about security? yes Kenn: besides, security is not just for humans. It's also to allow the system to safely reason about itself okay have you heard of aspect-oriented programming? if things can change in your back, you can make no assumption, hence, no advanced metaprogramming Kenn: sure Kenn: re-read my http://tunes.org/~fare/articles/ll99/index.en.html any taker to write the Scheme compiler? Fare: I'd write it maybe recruit someone on comp.lang.lisp or comp.lang.scheme? but I have not finished SICP Kenn: it's not such a hurry so then what are you doing? Kenn: are you doing all the exercises? Fare: yes writing an article, with the hope of submitting it to popl what is pop1 abi: popl? somebody said popl was at http://www.cs.wisc.edu/~reps/popl00/ Fare: the TUNES HLL will have an implicit operating system but how is the operating system internal to the HLL to allow reflection? The statement is correct, but can you repeat the question? how can I manipulate the OS within the HLL? Build the system out of the building blocks that it defines. but what if I want to use different blocks? like bricks? or stones A different terminology? yes Then you need to define what the new terminology means in terms of the old. There is no magic. what if you can't you're fucked and have to start all over ! 11:40am okay. how is migration different from meta-translation? Kenn: Linux is written in C. How can you manipulate Linux from Linux? Fare: is the Scheme compiler for use on Retro? Kenn: it doesn't exist yet It's not a problem to guarantee that you can represent one terminology in terms of another - the problem is that you may need to "add stuff" into the new terminology in order for it to work. -:- SignOff binEng: #TUNES ( ) but when it does, yes, I intend to try put it on Retro Fare: well, if we only use C, can we manipulate every part of it? i.e. not using inlined asm -:- SignOff AlonzoTG: #TUNES (Read error to AlonzoTG[209-122-253-198.s198.tnt8.lnh.md.dialup.rcn.com]: Connection reset by peer) Kenn: no, I wanted to show that there was no problem with manipulating a system from itself Kenn: I don't mean that Linux does it in the bestest possible way reflection? i heard reflection was a property of a system that can refer to itself and manipulate its state or rough on my brain For instance, linux requires you to reboot and loose all process state so as to modify the kernel LISP is _much_ easier to manipulate than C but Linux still has axiomatic code axiomatic? you mean imperative, perhaps? fixed must be accepted fundamental on TUNES, you could replace components on the fly because there will be an interface for it okay, so I can dynamically update the garbage collection algorithms? look this interface is the subject of my article Kenn: sure a meta-object protocol? Kenn: if the system has the right interface, you can change the GC algorithm (although this may still be a slow operation) (not something you'd do everyday) (unless you like that kind of thing) (sure) call it a MOP if you like calling things MOP sure confers them some hype value and can you manipulate the interface itself? is there a meta-interface? what do you call meta-i*f*? There appears to be a compromise between how complicated the compilation on something is and how often the you need to recompile it. something that manipulates the interface (cleanly) tmf: sure the interface is contextual right? tmf: Linux forces a static compromise on you tmf: Tunes cannot remove the compromise, but allows you to dynamically adapt it to tailor your needs speaking of Linux, I like Windows95 better contextual? Kenn: the things I do on linux, I cannot dream of ever doing them on win95 In TOOL, it may prove efficient to use a minimalistic compilation scheme ( i.e. object footprint only represent state and not structure ) ...