00:00:00 --- log: started forth/02.07.09 00:02:25 --- quit: ChanServ (carter.openprojects.net irc.openprojects.net) 00:02:44 --- join: ChanServ (ChanServ@services.) joined #forth 00:02:44 --- mode: carter.openprojects.net set +o ChanServ 00:04:33 --- quit: ChanServ (carter.openprojects.net irc.openprojects.net) 00:04:35 --- quit: n_ (carter.openprojects.net irc.openprojects.net) 00:04:35 --- quit: dsmith (carter.openprojects.net irc.openprojects.net) 00:04:36 --- quit: goshawk` (carter.openprojects.net irc.openprojects.net) 00:04:36 --- quit: Soap- (carter.openprojects.net irc.openprojects.net) 00:04:36 --- quit: rob_ert (carter.openprojects.net irc.openprojects.net) 00:04:36 --- quit: skylan (carter.openprojects.net irc.openprojects.net) 00:04:36 --- quit: cleverdra (carter.openprojects.net irc.openprojects.net) 00:04:36 --- quit: Fractal (carter.openprojects.net irc.openprojects.net) 00:05:20 --- join: ChanServ (ChanServ@services.) joined #forth 00:05:20 --- join: Fractal (pamixtbw@h24-77-171-228.ok.shawcable.net) joined #forth 00:05:20 --- join: rob_ert (~robert@h137n2fls31o965.telia.com) joined #forth 00:05:20 --- join: goshawk` (goshawk@panix1.panix.com) joined #forth 00:05:20 --- join: Soap- (~flop@202.0.42.22) joined #forth 00:05:20 --- join: skylan (sjh@Sprint270.tbaytel.net) joined #forth 00:05:20 --- join: dsmith (~dsmith@199.107.189.213) joined #forth 00:05:20 --- join: n_ (adlez@pcp01518417pcs.reding01.pa.comcast.net) joined #forth 00:05:20 --- join: cleverdra (julianf@0-1pool37-198.nas2.florence1.sc.us.da.qwest.net) joined #forth 00:05:20 --- mode: carter.openprojects.net set +o ChanServ 00:05:37 --- quit: ChanServ (carter.openprojects.net irc.openprojects.net) 00:07:49 --- join: ChanServ (ChanServ@services.) joined #forth 00:07:49 --- mode: carter.openprojects.net set +o ChanServ 00:08:12 --- quit: ChanServ (carter.openprojects.net irc.openprojects.net) 00:08:22 --- join: ChanServ (ChanServ@services.) joined #forth 00:08:22 --- mode: carter.openprojects.net set +o ChanServ 00:09:57 --- quit: goshawk` (carter.openprojects.net irc.openprojects.net) 00:09:57 --- quit: Soap- (carter.openprojects.net irc.openprojects.net) 00:09:57 --- quit: skylan (carter.openprojects.net irc.openprojects.net) 00:09:57 --- quit: rob_ert (carter.openprojects.net irc.openprojects.net) 00:10:07 --- join: skylan (sjh@Sprint270.tbaytel.net) joined #forth 00:10:07 --- join: Soap- (~flop@202.0.42.22) joined #forth 00:10:07 --- join: rob_ert (~robert@h137n2fls31o965.telia.com) joined #forth 00:10:07 --- join: goshawk` (goshawk@panix1.panix.com) joined #forth 00:15:53 --- quit: cleverdra ("Leaving") 04:32:53 --- quit: dsmith (Read error: 110 (Connection timed out)) 05:17:18 --- join: dsmith (firewall-u@cherry7.comerica.com) joined #forth 05:56:28 --- join: Etaoin (~david@216.251.197.132) joined #forth 07:24:24 --- join: cleverdra (julianf@0-1pool37-12.nas2.florence1.sc.us.da.qwest.net) joined #forth 07:51:33 --- join: kc5tja (~kc5tja@ip68-7-162-51.sd.sd.cox.net) joined #forth 07:53:38 * kc5tja is away: shower 08:23:16 * kc5tja is back 08:35:39 --- join: I440r (~mark4@1Cust172.tnt3.bloomington.in.da.uu.net) joined #forth 08:36:20 --- join: futhin (~thin@h24-64-175-61.cg.shawcable.net) joined #forth 08:41:38 OK, going to work now. 08:41:50 --- quit: kc5tja ("THX QSO ES 73 DE KC5TJA/6 CL ES QRT AR SK") 08:42:22 i440r 08:42:36 --- join: XeF4 (uvqlwx@12-245-116-85.client.attbi.com) joined #forth 08:42:40 i440r: i think you should write an outline for your tutorial 08:44:17 hi futhin! 08:44:28 yea that would help 08:44:47 it would help me a lot :P 08:44:52 heh 08:44:53 with editing.. 08:45:03 ive got the windowing thing almost 100% done now 08:45:04 unfortunately my urge to change everything is huge! 08:45:07 :P 08:45:13 just need to get pull down menus sorted out 08:45:16 well not that huge 08:45:41 but i do want it to be more easily readable 08:46:04 i'm fussy about that ;P 08:46:13 i440r: good to hear about the windowing thing 08:46:15 so am i 08:46:35 mouse control isnt built into it yet and i might not add it 08:46:41 undecided 08:46:47 nahh 08:46:52 mouse is lame :) 08:47:01 just implement keyboard stuff to control the windows 08:47:17 you can have overlapping windows NP but its easier if you dont heh 08:47:33 its not quite as fast as ncurses is at display updates tho 08:47:33 if you add mouse control, make sure that you can still do _everything_ without mouse 08:47:35 i440: is the tui in the latest isforth? 08:47:38 and thats confusing me alot 08:47:39 I enjoy using the mouse in console. 08:47:47 xef4 no - you wanna see it ? 08:47:52 rob_ert: mouse is slower 08:48:27 I440r: sure. An ms serial mouse driver is ~5 lines, so why not. 08:48:55 * futhin has some good ideas how to make a gui fully keyboard accessible so that one can operate in a gui at fast speeds from keyboard 08:49:18 it's faster overall, for one thing, never having to take your hands of the keyboard to touch the mouse saves some time 08:49:24 and there are lots of other advantages 08:49:27 --- join: Fare (fare@samaris.tunes.org) joined #forth 08:49:34 VOICE RECOGNITION! 08:49:34 hi fare! 08:49:35 scripting 08:49:36 etc 08:50:04 xef4 adding mouse support will slow down EVERYTHING 08:50:04 (+20 lines to deal with the brilliant foresight of the Unix serial port interface design) 08:50:21 i440r: i want you to create an outline and/or a table of contents.. and break up the tutorial with headings.. 08:50:22 you tell the console to send you mouse reports every time its moved or when a button is pressed 08:50:36 and you get them as ansi escape sequences from stdin 08:50:45 ALOT of shit thrown at you all at once 08:50:50 Gakuk! 08:50:53 I440r: Linux 2.2 seems not to obey those escape sequences. Xterm does, but I don't run X 08:51:00 futhin that works for me 08:51:21 XeF4: which sequences? to get mouse input? 08:51:27 XeF4: there's libgpm (ugh) 08:51:44 xef4 no - not "you send the terminal some escape sequcences" but "the terminal sends you escape sequences" every time the mouse moves 08:52:00 fare isforth will not be using ANY libraries heh 08:52:04 at all 08:52:05 I440r: you send the terminal some escape sequences to turn on mouse reporting 08:52:06 remember??? :) 08:52:22 xef4 yes and every terminal uses different sequences 08:52:26 almost 08:52:40 I440r: but is it forbidden to peer at the sources for libgpm 08:52:56 and the report sequences don't work for too-large terminals 08:52:59 fare its not allowed ? 08:53:09 nobody said that i440r is having trouble coding the mouse control 08:53:16 fare ohh no its not forbidden but i dont need to, i already know how to do mouse stuff heh 08:53:18 i440r said he was undecided in having mouse control in the first place 08:53:41 dns me 08:53:44 ftp in 08:53:50 tui.tgz is the code 08:54:05 i modified structs.f slightly 08:54:17 i put the new structs.f in the tgz there 08:54:28 xef4: i440r wanted to send you his windowing stuff because his display update isn't fast and it's confusing him 08:54:31 but keep the old one too 08:54:44 futhin no i know how to speed it up too hhe 08:54:50 ew 08:54:50 futhin: I asked him to send. 08:55:07 but i wont be able to speed it up till i can do CODED definitions 08:55:09 [09:47] xef4 no - you wanna see it ? 08:55:09 [09:47] rob_ert: mouse is slower 08:55:09 [09:48] I440r: sure. An ms serial mouse driver is ~5 lines, so why not. 08:55:33 anyone wanting to see the tui its in the root of my ftp on my current ip 08:55:46 that's a little misleading, xef4 seems to imply he's expecting to recieve mouse code 08:56:10 futhin no heh your the one getting confused here :) 08:56:28 heh 08:56:41 naw, i'm just telling you what it implies :P 08:57:12 Actually I didn't see i440r imply he was expecting anything. 08:58:04 tui.tgz: Permission denied. 08:58:13 oopts 08:58:14 hang on 08:58:32 ok try again heh 08:58:44 whee 08:58:52 the menu.f file is in progress 08:59:00 window and screen are mostly finished 08:59:13 foo.f is a bullshit example i am using to debug with 09:00:23 bah, you should've called it blah.f :P 09:01:05 fooblah :) 09:01:06 all real coders use blah for temp files ;) 09:01:45 i use foo, blah, foobar, fooblah 09:03:22 i have blah and bleh 09:04:45 zip zap zop zup 09:05:03 a b c d 09:05:59 if its directories, i and there's more than one, then i'll go 1 2 3 4 09:06:39 further evidence that Dr. Seuss was a real coder 09:07:06 but for files, blah and bleh are sufficient, if i run out i'll start giving them more meaningful names :P 09:09:42 i440r: you asked me to create content for your webpage.. i uploaded it into the incoming directory 09:09:49 i440r: it's REALLY basic 09:09:52 heheh 09:11:08 futhin i got a guy sorta doing the page 09:11:21 i440r: you asked me for content 09:11:26 the content needs work and he is only just learning html but he is making progress 09:11:27 don't you remember? :) 09:11:27 oh 09:11:30 rite content 09:11:31 yea 09:11:32 heh 09:12:14 i440r: it's probably easier for you to write some content first 09:12:19 i know 09:12:26 but i hate coding ascii :P 09:12:30 heheh 09:12:35 hey that's a good idea 09:12:40 just code it in forth 09:12:44 with lots of comments 09:12:54 and we'll just put it up in text form 09:12:57 :D 09:13:17 s/we'll/you! 09:17:55 ahh 09:18:01 i see that you are already coding it! 09:18:03 good work 09:18:13 err already coding what ? 09:18:13 heh 09:18:18 im not coding shit rite now :P 09:18:45 coding the content for the webpage in forth which you'll put up in text/ascii form!! 09:20:23 i'm free today to do something productive 09:20:29 coding in forth or something 09:20:45 and damned if i do, and damned if i don't! :P 09:21:47 write me that outline 09:22:37 okay 09:22:40 how big is the tutorial 09:22:42 what is the aim 09:23:04 --- join: kc5tja (~kc5tja@stampede.org) joined #forth 09:23:40 wb kc 09:23:56 start with basic forth tutorial stuff then lead into isforth specifics 09:24:14 and what audience is the forth tutorial aimed at? 09:24:34 err forth coder wannabe's ??? heh 09:24:42 and ppl who know forth but want to learn isforth 09:25:22 what about skeptics? 09:25:27 people who think forth is crap 09:25:29 C coders 09:25:34 theres no convincing them 09:25:38 they are beyond help 09:25:50 i started writing a forth tutorial a bunch of months ago 09:26:03 aimed at C coders and other losers 09:26:16 irrational people, stuff like that ;P 09:26:46 Where's the tutorial? I'd like to see it. 09:26:59 i only wrote a page heh 09:27:01 it's not finished 09:27:06 2 pages 09:27:16 it was turning into a 10 page essay 09:27:17 The best way to convince a C coder is to make things as pragmatic as possible (hence, it's not specifically targetted towards C coders, but rather, towards people who just want to do something useful). 09:27:28 yes 09:27:37 i was going into details of why forth is useful 09:27:43 but it's awfully tricky 09:27:48 If you write your tutorial so that it singles out languages, then you get a reputation for being a ball-buster and troublemaker. 09:28:00 to be readable, and to do a decent job of covering all the various advantages 09:28:10 yeah, the bad-boys of tutorial writers! 09:28:13 and to explain the advantages clearly and concisely 09:28:23 it's pretty tricky 09:28:33 to write an exceptional tutorial 09:28:46 futhin - I'd really like to see it, even if it's unfinished. 09:28:48 * kc5tja nods 09:28:54 that thoroughly covers all the really complex concepts of forth 09:29:08 what complex concepts? 09:29:14 even though forth is "simple" it is also full of highly complex concepts and paradigms 09:29:30 The simpler the language is, the more complex it becomes. 09:30:17 OK. What complex concepts? 09:30:18 complex concepts = all the concepts that C coders or "pragmatics" would disbelieve and discount at first glance if the tutorial doesn't carefully explain the forth concepts 09:30:32 it's just complex in relation 09:30:38 at the moment i can't think of the complex concepts 09:30:40 WHAT COMPLEX IN RELATION CONCEPTS? 09:30:41 of an example 09:30:43 Oh, OK. 09:30:46 lol 09:31:03 * futhin goes to look for his tutorial 09:31:31 A complex concept would be literally adding support for continuations and backtracking to the language in terms of the language itself. 09:32:09 It requires direct manipulation of the return stack in non-portable (but *super* easily fixed) ways. 09:32:23 C programmers will absolutely balk at such a concept, while in Forth, it's openly embraced. 09:32:25 a complex concept is how forth _is_ the language, the interface, the compiler 09:32:36 a complex concept is how forth is reflective 09:32:37 etc 09:32:51 It's reflective without actually supporting it. 09:33:00 yeah 09:33:12 hence i said "how" ;P 09:33:31 How is Forth "reflective"? 09:33:43 kc5 - ah. 09:34:09 The amount of reflection is dependent on the Forth in question, but one thing is quite common: you can always find a way to query the Forth environment for its list of words. 09:34:27 What do you mean when you say 'reflective'? 09:34:57 when i say reflective i mean that you can modify the language within the language easily, you can change the language freely into anything 09:35:00 The program is aware of its surroundings, is the best way I can describe it. It can detect and manipulate its surrounding program code. 09:35:43 OK. 09:36:00 * cleverdra notes http://forth.sourceforge.net/algorithm/backtracking/index.html 09:36:01 Logically, reflection is the logical opposite of encapsulation. 09:37:04 the forth paradigm is foreign to C coders 09:37:17 the paradigm is quite different than what they are used to.. 09:37:57 i assume that i accepted forth much more readily because i had learned assembly programming previously (a new paradigm).. 09:38:00 * kc5tja is really, really, really starting to come more and more to grips with exokernel philosophy. It's really in line with Forth concepts, I believe. 09:38:12 it's a lot easier to explain forth from assembly language 09:38:19 rather than from C 09:38:36 (refering to how forth behaves, macro assembler, etc) 09:38:45 futhin: Absolutely. This is why I'm very upset at colleges NOT teaching assembly language anymore. It's an absolute requirement if you ever want to be a free thinker in computer science. 09:39:13 my university teaches assembly but it's some pre-chewed version and the prof is anti-asm 09:39:34 it's in mips too, not that "useful" 09:39:54 i think computer programmers should serve an apprenticship periopd where the are not allowed to code in ANY language other than assembler for the first 15 years 09:39:56 exokernel? what's in it? 09:39:58 minimum 09:40:16 pre-chewed? 09:40:20 i440r: heheh, bit extreme 09:41:03 kc: tell us about the exokernel philosophy :) 09:44:04 In an exokernel, the only thing that sits in kernel space is a minimal kernel that provides resource protection. It does not, however, provide any form of management. 09:44:55 --- join: Mongrel (~ant@ool-435249eb.dyn.optonline.net) joined #forth 09:44:56 A "Linux" application, for example, would be an application that links against a user-space Linux library (liblinux.so for example), and runs under the exokernel entirely in user space. 09:45:07 hi mongrel 09:45:26 fare: yeah, school has a tendency to package up knowledge into little pre-chewed bits which retain a bare minimum of information, and then the school transmits it poorly to the students and hopefully with lots of bias 09:45:28 Performance of exokernels is competitive with traditional monolithic designs, but you have the relative modularity of microkernels, and a flexibility which *neither* possess. 09:46:36 hm? what's the diff between microkernel and exokernel? 09:46:55 i don't see the speed advantages of the exokernel like the monolithic kernel has 09:47:28 Microkernels puts the functionality in user-space server processes, which are self-standing processes in their own right. You request services from them using interprocess communications channels (message passing, pipes, sockets, whatever). 09:47:31 Exokernels don't do this. 09:48:03 With exokernels, the software to manage resources is literally linked right up against the application -- requesting a "system service" is a mere subroutine call (not even a kernel trap!). 09:48:27 What the exokernel DOES provide, however, is protection. 09:48:55 futhin - have you found your tutorial? 09:49:13 For example, if disk blocks 2 through 5 are used by a file owned by you, the exokernel will not allow my application (which may not even be running a Unix variant) to touch those blocks. 09:50:13 Exokernels expose all hardware and kernel resources at the lowest possible level while still maintaining protection. 09:50:18 kc5tja: is the library in user-space in system-space, protection-wise? 09:50:23 Microkernels tend to hide these resources behind interfaces. 09:50:30 User-space. 09:50:58 who schedules disk accesses? 09:51:17 how is the ownership map represented? 09:51:22 who manages it? 09:51:47 The exokernel guards access to the disk controller, but it's usually first-come-first-serve. 09:52:08 \quit 09:52:11 --- quit: Mongrel ("BitchX: often imitated, never duplicated!") 09:52:52 Ownership maps aren't natively represented in any given format. Individual protected entities are represented with downloaded code (e.g., code that is dynamically compiled into kernel space). These are called "predicates." 09:52:59 Individual applications manages their own resources. 09:53:56 kc5tja: have you read "LAMBDA, the Ultimate capability" ? 09:54:01 cleverdra: yes.. it's only one page, it's just the bare intro, hasn't gotten into any juicy details.. 09:54:09 aka "W7, a security kernel based on the lambda-calculus" 09:54:13 Thus, what basically happens is the exokernel grows and shrinks (at run-time) to accomodate the resource protection needs of individual applications and their library OSes. 09:54:44 Fare: No, because I do not understand Lambda calculus, and I really don't care to learn about it. 09:55:32 kc: there are nice tutorials, and you're missing something 09:55:48 who manages this downloadable code? 09:56:49 Fare: Again, the applications manage their own resources. The exokernel only provides protection. 09:57:13 That means if they install a predicate into the OS to guard against a certain set of disk blocks, then that application is responsible for managing it too. 09:57:41 The exokernel enforces the protection aspect of it: it cannot overlap with another predicate. 09:58:06 (ooh, and if the task quits or dies, it's predicates are removed as well, of course. Protection without basic garbage collection is useless) 09:58:18 kc5: I read the W7 paper not knowing uch about the lambda calculus, and it was still understandable 09:58:36 IOW the protection isn't effective against evil/buggy predicates 09:59:26 (which already narrows protection problems) 10:00:08 Not unless all the other predicates are also evil/buggy. 10:00:21 The exokernel will detect overlapping predicates, and will reject all of them. 10:00:30 cleverdra: you still want to see it? :) 10:00:58 If disk block 56 appears in multiple predicates, all predicates except the first will be flat out rejected (they won't even be considered for installation into the kernel) 10:01:34 how can it detect overlapping predicates? 10:02:00 That's exokernel specific. 10:02:11 If you want to know more about how Xok does it, I recommend reading Xok's papers. 10:02:28 IOW ? 10:02:38 is Xok still being developed? 10:02:41 In Other Words 10:02:57 One possible method would be to maintain a bitmap or something, and as each predicate is installed, it marks their associated bits (where each bit maps to a disk block). 10:03:11 Before installing a predicate, a check is made to ensure all mapped bits are cleared first. 10:03:34 Yes, Xok is still being developed. So is Aegis as far as I know. 10:04:25 what are xok and aegis ? 10:04:34 I440r: exokernel designs 10:04:35 exokernel oses 10:04:40 i440r: stuff done at the MIT 10:05:18 * kc5tja is seriously thinking writing Dolphin to use an exokernel instead of microkernel. 10:05:29 The advantages are just plain too enticing. 10:05:52 it's rather than microkernel is so dumb a concept, that anything is better, including exokernel 10:06:34 Well, the fact is, when you have several applications running with a single libOS, the exokernel approach is semantically equivalent to a full-block monolithic kernel design. 10:07:14 The nice advantage is that you can run "multiple OSes" concurrently with the exokernel approach (of course, you're not really doing that, but that's the end result as far as the user and applications are concerned) 10:07:17 kc: where along the development is Dolphin? you coded a target compiler or something? but you have not yet started on the actual os and the kernel? 10:07:45 Each installed resource protection predicate is like a fine-grained kernel module (usually expressed as a set of data structures instead of actual program code, for protection purposes). 10:08:39 futhin: The target compiler is still under construction. It's not yet finished, but it is producing 16-bit code. When I get a self-standing version of FS/Forth 16-bit running, I'll use it to make a 32-bit version, and from there, start writing the Dolphin (exo?)kernel. 10:10:25 I really like the concept of using an exokernel for another reason: I can write most of Dolphin under Linux, or some other environment that has reasonable fault tolerance. 10:10:49 The "exokernel" basically ends up being an emulation layer, which makes things much easier to debug. 10:12:22 My friend wanted me to consider making Dolphin a persistent OS too, but I haven't a clue as to how to go about doing that. 10:12:39 Every orthogonal persistent system I've seen really is fundamentally incompatible with what I want out of the OS. 10:13:14 * kc5tja may consider doing another OS which is based on Dolphin but which has orthogonal persistence in the future, but not right now. 10:13:45 can't an orthogonal persistent os still mimic a standard os's filesystem, etc 10:14:12 Yes, but the reverse is not true. 10:14:24 Plus, an orthogonally persistent OS has *major* problems dealing with removable media. 10:14:25 --- join: Herkamire (~jason@wsip68-15-54-54.ri.ri.cox.net) joined #forth 10:16:24 isn't that a solvable problem? 10:16:59 --- quit: XeF4 ("pois") 10:17:02 Nope. 10:17:03 hello herk 10:17:22 Nobody, and I mean nobody, has yet offered a truely satisfactory solution. 10:17:54 you were telling me about the problem, but i thought it should be solvable.. 10:18:03 The best they can do is treat the system's harddrive as a virtual memory store, and physically copy program images onto the harddrive -- a waste of space, especially if you're running software from another harddrive already. 10:19:08 I mean, the whole purpose of having additional volumes is to expand the virtual address space, not shrink it. 10:19:25 And that brings to mind virtual address space management with multiple volumes. 10:19:42 kc5tja: how is it worse for an orth persistent system than for another one? 10:20:05 if you remove whatever you mounted as / from a unix system, you lose 10:20:18 it isn't wors 10:20:19 You can't umount / in Unix. 10:20:25 Let me explain, please. 10:20:30 Don't just jump to conclusions. 10:20:49 Fuck, I forgot where I was going now. 10:20:53 See, I hate it when people do this. 10:20:56 * kc5tja thinks. 10:21:12 so the goal is to be able to add another harddrive to the computer and then turn the computer back on and it can handle the extra volume with no problems 10:21:22 futhin: Basically yes. 10:21:33 Suppose I have two boxes, each with a 250MB ZIP drive. 10:21:53 so why can't the start up treat the memory dynamically or something.. 10:21:55 you want to physically swap memory? 10:22:00 1. When I run an instance of a program, how do I tell the application to run in the ZIP's physical address space? 10:22:34 2. Assuming that problem is solved, how does the second host interpret the memory image on the ZIP? They may not necessarily appear in the same physical memory range. 10:23:13 The only way to alleviate both of these problems is to have some sort of API where the application can read and write memory regions as a separate address space from itself. 10:23:31 That secondary address space is essentially what a file is. 10:24:01 That is the only (efficient) way to handle the problem. Sub-optimal, yes, but to date, no other alternative has ever been proposed. 10:25:00 you could have system contexts whereby applications running in that context trust the ZIP disk to be in, and hang when it's not 10:25:18 That still doesn't tell me how to instruct the program to run in that context. 10:25:26 (the context controller, that must not trust the same condition, might get an exception in that case) 10:25:38 that's just a matter of additional syntax in your shell 10:25:54 * kc5tja shakes his head -- that won't do. 10:26:12 If you can specify one context, you must be able to specify multiple contexts. 10:26:22 using-memory-from ZIP1 my-program 10:26:23 Otherwise, "copy" type programs will not be possible. 10:27:14 copy programs are obviously at the meta-level, just like the context controllers, with respect to the store. 10:28:20 i.e. you can't do low-level operations on a disk you mounted 10:28:20 * kc5tja disagrees with the whole concept. 10:28:43 just like mounting in unix, except each process may have a different mounting context 10:29:01 Well, each process gets a different VM swap space. 10:29:21 so what's the problem with a different mount swap space? 10:29:30 It's unnecessarily limiting. 10:29:38 kc: i wouldn't worry about orthogonal persistence resembling too closely the old filesystem way.. as long as the orthogonal persistence advantages are still there.. 10:29:38 how that, limiting? 10:30:03 futhin: You still have to manage your data. 10:30:41 no you don't 10:30:41 Sure you do. 10:30:41 Absolutely you do. 10:30:41 you just add a mount somewhere in your initialization file, and there you are 10:30:41 What if I want to edit a text document I wrote in Notepad under a more sophisticated word processor now? 10:30:50 Am I expected to hand cut-n-paste it? 10:31:08 and every time you need a *specific* mount for a program, you wrap your program in a function that launches it in the proper context 10:31:28 Such a system is becoming increasingly more complex. 10:31:46 kc5tja: how is that different from moving files between Wordperfect and Word? 10:31:46 Everytime you need a wrapper, that's extra complexity. 10:31:57 Because the "file" exists in the program's address space! 10:31:58 kc5tja: of course, when you exchange data, you need a common file format 10:32:05 so what? 10:32:25 you can copy data from address spaces to another one. 10:32:41 That's right -- you need a common file format. That defeats the purpose of orthogonal persistence, yes? The whole idea behind it is to completely eradicate the traditional filesystem. 10:32:44 copy-to ZIP2 (somedatainZIP1) 10:33:09 with orth persistence, you don't need a common *file* format, but a common *datatype* 10:33:22 i.e. you don't have to worry about marshalling 10:33:27 you only worry about structure 10:39:58 Any translation from internal to external representation is still marshalling. 10:40:21 what external representation? 10:40:21 you don't need it 10:40:21 The common datatype. 10:40:21 it's not marshalling 10:40:21 A text editor will not just edit a flat buffer of bytes. It'll break it up into lines at the very least, to facilitate rapid response. 10:40:21 your question is about as stupid as the one some MP asked Babbage: "if we ask the wrong question, will it give the right answer?" 10:40:21 Yet, when transfering it to another program, you want that flat buffer of bytes. 10:40:21 if you run a program with a wrong parameter, it won't work 10:40:21 My question isn't as stupid, it points out a fundamental inconvenience of the otherwise very convenient system. 10:40:21 it's not an inconvenience of orth persist 10:40:21 it's an inconvenience of just any computer system 10:40:21 a type mismatch is a type mismatch in just any computer language or system 10:40:22 Now you're introducing language restrictions. 10:40:22 More complexity. 10:40:22 if you S" sdf" 1 2 F+ you lose 10:40:22 Well, no kidding. 10:40:22 But that's not orthogonal persistence! 10:40:22 if I edit a file, and move it to another editor, I convert it to character string first 10:40:22 Marshalling. 10:40:22 (or to some object that obeys some proper protocol) 10:40:22 no 10:40:22 Yes 10:40:22 it could be a "virtual" string 10:40:22 The definition of marshalling is to convert an internal data representation into an external, well-known format. 10:40:22 that just obeys the string protocol 10:40:22 no need to marshall anything 10:40:22 It does NOT matter if it's being stored in a file or in RAM> 10:40:22 no. 10:40:22 the definition of marshalling is to convert internal data representation into a stream of bytes. 10:40:22 Well, hell, /dev/random does that just fine. 10:40:22 But that doesn't make it marshalling. 10:40:22 Saving a spreadsheet document also does that, but unless you know what *FORMAT* that stream of bytes is in, it may as well have been from /dev/random. 10:40:22 you don't need copy or move anything around if you have a proper wrapper. 10:41:30 and if you need to copy something from a persistence space to another, that's just a one instruction that won't need do anything more with orth persist than with explicit persist 10:41:30 Sure. But if you want a program that actually runs with any respectible performance, you'll not use any wrappers at all until they're necessary. 10:41:30 * futhin is thinking that both of you guys have slightly different ideas of what orthogonal persistence IS 10:41:30 you're talking rubbish 10:41:30 * kc5tja laughs 10:41:30 Whatever. 10:46:17 the explicit persistence system is the one that uses a low-level wrapper 10:46:17 a forced one 10:46:17 around all applications 10:46:17 You can think what you want, but to date, you haven't given me anything to convince me otherwise about orthogonal persistence. 10:46:17 with orth persistence, most applications directly use objects in memory 10:46:17 and it's only when you have a type mismatch that you must convert things 10:46:17 Yes, I know full well what the differences are. 10:46:17 which might not be costly 10:46:17 But you yourself admit that you need common data formats. 10:46:17 no, not formats 10:46:17 types 10:46:17 big different 10:46:17 BIG difference 10:46:17 Not when the program manipulates the memory image directly! 10:46:17 what difference does that make? 10:46:17 all applications, text files, etc.. are objects 10:46:17 so what? 10:46:17 * kc5tja doesn't know how to respond to that. 10:46:17 and the objects have wrappers 10:46:17 as if explicit persistence was magic 10:46:18 (i'm talking out of context here) 10:46:18 I cannot believe you just asked that question. 10:46:18 why do they have wrappers? 10:46:18