00:00:00 --- log: started forth/04.09.04 00:03:50 --- quit: Sonarman ("'nn") 00:06:16 re 00:06:24 Sorry -- talking to the new roommate. 00:08:07 sup kc 00:08:09 hows life 00:10:11 It's life, Jim. But not as we know it. 00:40:24 lol whats that from 00:55:01 Saturday Night Live skit on Star Trek. 00:55:12 Anyway, I'm off to bed. 00:55:24 --- quit: kc5tja ("THX QSO ES 73 DE KC5TJA/6 CL ES QRT AR SK") 00:56:35 --- quit: warpzero (Remote closed the connection) 01:07:10 --- join: warpzero (~warpzero@dsl.103.mt.onewest.net) joined #forth 02:03:45 --- join: ASau (~root@217.16.31.100) joined #forth 02:03:53 Dobryjj den'! 02:06:29 Privet, ASau :) 02:13:33 God dag, Robert! 02:13:40 What's new? 02:19:22 Not too much, really 02:28:41 --- join: Topaz (~top@spc1-horn1-6-0-cust117.cosh.broadband.ntl.com) joined #forth 02:33:33 Hi 02:35:40 mjello 03:11:09 --- quit: fridge ("rock out with your cock out") 04:05:29 --- join: crc (crc@76-pool1.ras11.nynyc-t.alerondial.net) joined #forth 04:09:40 Dobry den, crc! 04:09:49 Anything new? 04:11:33 RetroForth 7.3 Beta #1 is out 04:11:43 Updating the documentation right now :-) 04:12:08 I've been playing with the new system call interface too 04:12:42 I changed the conditionals... 04:12:50 Instead of = < > <> if 04:12:58 I now have =if if <>if 04:13:05 (Like in herkforth) 04:13:39 And it has counted loops (repeat/until) 04:14:16 * crc is also working on support for 64-bit integers, but that's still very buggy 04:32:48 Sorry, I've been out. 04:33:12 I can share double width arithmetics. 04:34:08 I've worked out addition and subtraction 04:34:22 The problem is figuring out how to add them cleanly to RetroForth 04:34:24 Really, that's not good if you excluded old conditionals. 04:34:31 Why? 04:34:46 Too many code relies on them. 04:35:00 'sed s/= if/=if/' 04:35:15 * crc will provide a conversion script 04:35:26 When you exclude them, you have to rewrite old code from conjunctions and disjunctions to branching. 04:35:36 ??? 04:35:59 "dup 0= swap X = or" ? 04:36:21 I've never used code like that 04:36:34 And it's trivial to add it back if you need it 04:36:51 Well, I know that. 04:37:23 But this makes to waste time and requires additional efforts. 04:40:37 It only requires more effort if you need to use code like your example 04:41:49 This is more compact, cleaner to read, and faster :-) 04:42:07 This case is more usual, than yours. 04:42:17 Early optimization is not good. 04:42:23 Why? 04:42:39 This is more compact only in the simplest cases. 04:42:54 How so? 04:43:40 When you have to write disjunction, you have to negate two prior results and write two branching constructions. 04:43:49 ?? 04:43:54 --- quit: tgunr (Read error: 54 (Connection reset by peer)) 04:44:00 See: 04:44:24 cond1 cond2 or if do-this then 04:44:27 against: 04:44:48 cond1 not if cond2 not if do-this then then 04:45:14 Even when you have shorter "if" it is no simpler. 04:45:37 --- join: tgunr (~davec@vsat-148-63-4-107.c001.g4.mrt.starband.net) joined #forth 04:46:43 What you have to do in case of exclusive disjuction makes headache. 04:47:35 The simplest solution is low level hack. 04:49:01 It is using substraction. Like in ancient forthes. 04:52:09 What should "IF" check for then? 04:53:28 The old "IF" was: 04:53:31 test eax, eax 04:53:34 lodsd 04:53:36 jz ???? 04:53:58 --- quit: Topaz (Remote closed the connection) 04:55:56 (EAX=TOS, ESI=pointer to rest of stack. [ESI] = stack item below TOS) 04:59:27 In your example (cond1 cond2 or if do-this then) aren't you checking that cond1 and cond2 resolve to the same value? 04:59:53 So couldn't you do this?: cond1 cond2 =if do-this then 04:59:58 * crc is confused here 05:03:20 In the old "IF" (from quick tests), any value other than "0" was considered TRUE 05:03:52 I still fail to see how, for normal use, the new conditionals are inferior to the old 05:40:05 "OR" tests both values to be non-zero. 05:40:32 "=" tests both values to be equal. Eather zero or non-zero. 05:40:38 Either... 05:46:51 --- quit: tgunr (Read error: 104 (Connection reset by peer)) 05:48:33 --- join: tgunr (~davec@vsat-148-63-4-107.c001.g4.mrt.starband.net) joined #forth 05:48:58 So "OR" sets the "z" flag? 05:49:03 * crc was AFK 05:52:30 On x86, yes. 05:53:34 I see 05:53:51 But the old IF would do 'test eax, eax' invalidating that anyway... 06:00:39 Crc, trying to optimize do not eliminate general cases. 06:01:35 When one want to optimize, he'll find how to do that. 06:01:55 It's not just an optomization 06:02:02 It's an attempt to simplify out things 06:02:27 Simplification is optimization too. 06:02:50 Perhaps 06:03:03 A particular case of optimization. 06:03:08 But I *never* use conditionals apart from IF 06:04:12 I and many mathematicians use logical functions as usual functions. 06:04:44 I'm not aiming Retro at mathematicians though 06:04:55 That's also vote for logical true to be 1 and never -1. 06:06:20 I've never read Gibbs in English. 06:06:35 Once he said that "math is language." 06:08:20 The old IF would allow anything *except 0* as TRUE 06:09:27 For complex math, RetroForth isn't a good choice. It only supports 32-bit integers 06:13:58 --- join: Topaz (~top@spc1-horn1-6-0-cust117.cosh.broadband.ntl.com) joined #forth 06:19:28 32-bit integers are enough for complex math. 06:20:08 Either doubling width is applicable. 06:20:39 Software floating point numbers are not unusual also. 06:21:43 You also don't have a lot of operators ( + - * / mod negate or xor and not << >> ) IIRC 06:31:59 --- join: FlamingRain (~Ecoder@c-24-129-95-254.se.client2.attbi.com) joined #forth 06:32:45 --- quit: ASau (Read error: 104 (Connection reset by peer)) 06:37:09 --- join: ASau (~root@217.16.31.100) joined #forth 06:49:50 --- quit: crc ("Time for bed... Goodnight!") 06:59:56 --- join: Herkamire (~jason@h000094d30ba2.ne.client2.attbi.com) joined #forth 06:59:57 --- mode: ChanServ set +o Herkamire 07:15:16 --- quit: ASau ("leaving") 07:48:41 --- quit: warpzero ("Tried to warn you about Chino and Daddy Gee, but I can't seem to get to you through the U.S. Mail.") 07:51:21 --- quit: FlamingRain ("Leaving") 08:15:49 --- join: tathi (~josh@pcp02123722pcs.milfrd01.pa.comcast.net) joined #forth 08:19:59 --- join: kc5tja (~kc5tja@66-74-218-202.san.rr.com) joined #forth 08:20:00 --- mode: ChanServ set +o kc5tja 08:36:02 --- join: warpzero (~warpzero@dsl.103.mt.onewest.net) joined #forth 09:12:19 Getting ready for work. brb 09:12:24 See you 09:13:47 --- quit: qFox ("meh ircii's netsplit detection appearantly sux ^^") 09:21:48 --- quit: tathi ("leaving") 09:27:37 --- join: zruty (~chuck@64-121-15-14.c3-0.sfrn-ubr8.sfrn.ca.cable.rcn.com) joined #forth 09:34:54 back for now 09:38:01 Hello then 09:38:23 * Robert is reading an interesting book on computability and logics. 09:40:09 * kc5tja nods 10:08:12 * kc5tja is reading up on the Treadmill garbage collection algorithm. 10:12:40 * Herkamire is attemting to redesign his business website 10:16:34 web design is a funny business 10:17:17 essentially just art, although the best artists tend to make terrible web designers due to their completely inability to comprehend HTML/CSS 10:17:58 so they produce huge imagemaps, images for all the text in the site (annoyingly common), etc etc 10:19:11 Yep. 10:22:21 web design should be UI design, not just graphic design 10:24:39 yeah 10:24:43 except my mom. 10:24:48 she makes designs that work great on the web 10:25:41 but in many cases, the UIs are extremely simple 10:25:46 Man, I can't understand this Treadmill thing. 10:25:50 (most cases?) 10:26:14 I just don't see how keeping things in a circular loop would in any way improve the performance of garbage collection. 10:26:19 But I digress. I need to get to work now. 10:28:18 The sad thing is, UI design ideally is very spartan, but everywhere you look, there is some wanna-be beatnik trying to install custom shells on their computer, destabilizing the hell out of it as they install their fluffy-puff window decorations and what not. They want their websites to be just as "Visually Exciting!". 10:28:24 I give up. 10:28:26 Peace. 10:28:29 --- quit: kc5tja ("THX QSO ES 73 DE KC5TJA/6 CL ES QRT AR SK") 10:31:26 elegance in simplicity 10:32:05 i do like pretty pictures though 10:32:11 as long as they stay put ;) 10:32:21 my site is rather silly 10:32:28 http://www.ecs.soton.ac.uk/~jb1403/ 10:39:11 simple is actually good 10:39:23 ui design is "do the least thing that is recognizable and works" 10:39:31 the rest is polish 10:39:47 yeah, but eye candy is surprisingly influential 10:39:58 of course "works" has a big scope wrt accessability and L10N 10:48:16 Topaz: influential towards what? 11:00:57 well, people's first opinions 11:01:01 --- join: Sonarman (~matt@adsl-64-171-254-189.dsl.snfc21.pacbell.net) joined #forth 11:01:19 ie, a site which looks incredibly boring will often make people think the company is like that 11:07:44 --- quit: tgunr (Read error: 104 (Connection reset by peer)) 11:09:17 well, for my business booring is much better that cluttered or disorganized or confusing 11:09:35 --- join: tgunr (~davec@vsat-148-63-4-107.c001.g4.mrt.starband.net) joined #forth 11:09:43 I need to look reliable and helpful. Not cute 11:36:30 hi 11:39:51 --- quit: tgunr (Read error: 104 (Connection reset by peer)) 11:41:48 --- join: tgunr (~davec@vsat-148-63-4-107.c001.g4.mrt.starband.net) joined #forth 11:46:05 --- quit: zruty (Read error: 110 (Connection timed out)) 11:49:04 --- join: thefox (~fox@adsl-68-124-64-61.dsl.pltn13.pacbell.net) joined #forth 11:49:23 thefox, what's new? 11:49:40 hey does anybody here want to give me shell on a 64-bit machine? 11:50:00 been busy 11:50:24 i *think* my interpreter is now 64-bit clean, but of course it needs testing. string packing might have endianess issues 11:51:31 thhi thefox! 11:52:01 slava ill have isforth running in fbsd natvily today or tomorrow i think :) 11:52:09 cool 11:53:29 i'll start studying x86 asm around that time too :-P 11:53:30 I440r: yuriz (NASM developer) is in #nasm now, but says NASM development is pretty much very very dead. 11:54:17 i think i've figured out how to adopt my garbage collector to handling compiled code. 12:07:10 forth doesnt compile garbage code :) 12:07:11 hehe 12:07:17 actually humor aside 12:07:26 how can any code be garbage collected ? 12:07:32 I440r, it does if you redefine a word, and recompile it, and no longer want the old code to hang around 12:07:41 oh yea 12:07:50 well 'normal' forth doesn't allow that 12:07:51 thats not easy 12:07:53 since the dictionary just grows 12:08:06 at first i'll just have old code never be deallocated 12:08:08 its not a big deal anyway 12:08:12 because every single word that references it must be patched 12:09:29 I440r, are indirect relative jumps on x86 slow? 12:09:42 err prolly 12:10:02 im not even sure how to do a relative indirect jmp 12:10:06 heh 12:10:12 n/m 12:10:27 i know what u mean 12:10:39 nut x86 doesnt have that mode i dont think 12:25:45 --- join: kc5tja (~kc5tja@66-74-218-202.san.rr.com) joined #forth 12:25:45 --- mode: ChanServ set +o kc5tja 12:25:50 * kc5tja is now at work. 12:26:22 Hi :) 12:30:17 --- join: T0paz (jonny@spc1-horn1-6-0-cust217.cosh.broadband.ntl.com) joined #forth 12:31:59 what do you think of the new design I'm working on: http://jason.herkamire.com:8888/herkamire/w/index.html 12:32:18 I'm not entirely happy with it, but I like it better than the current: http://herkamire.com/ 12:34:04 --- quit: Topaz (Read error: 113 (No route to host)) 12:38:06 * kc5tja totally has an upset stomach right now. :( 12:38:09 * kc5tja checks 12:38:39 "The connection was refused when attempting to contact jason.herkamire.com:8888." 12:40:36 kc5tja! 12:40:48 slava! 12:43:38 I'm so well liked, that I apparently leave everyone speechless. :) 12:44:12 weird... dns failing completely 12:45:49 I can bring up your home page just fine. 12:45:54 I just can't access port 8888, it seems. 12:46:02 somehow not propagating 12:46:22 It *just* happened then. 12:46:24 the port 8888 url is my home computer 12:46:27 I am also confirming. 12:46:34 jason.herkamire.com is not propagating right somehow 12:46:36 Weird. 12:46:50 it resolves fine at home, but not From the server (going through normal nameservers. 12:47:00 What is its IP address? 12:47:04 my name server is responding fine to direct queries 12:47:13 24.218.95.147 12:47:19 dig: Couldn't find server 'ns1.jwoof.com': Name or service not known 12:47:28 That's probably one of the reasons. :) 12:47:34 "dig +trace" is your friend. :) 12:47:41 http://24.218.95.147:8888/herkamire/w/index.html 12:47:59 Connection is still refused. 12:48:23 "The connection was refused when attempting to contact 24.218.95.147:8888" 12:48:40 works 12:48:41 here 12:48:47 hmm, tis cute 12:49:00 dunno how it scales to displaying lots of data though 12:49:07 the kick game is now active... the next person to say anything will get kickbanned... (ops excluded) 12:49:07 SeaWorm is bottom feeder i am far from being superior.. just dont like idea that we have to call an array container vector (which might confuse any high school physics student who you algo) 12:49:07 --- l_oke_ sets ban on *!*@static-89-213.dsl.tht.net 12:49:07 <-- docjava has kicked SeaWorm from #java (Banned) 12:50:12 What the fuck? 12:50:24 This is why I stay away from mainstream IRC channels. 12:50:30 They're run by a bunch of retards. 12:51:05 --- quit: thefox () 12:51:22 T0paz: thanks. the sides get taller with more data. 12:51:42 Hmm...I still can't access the site here. :( 12:51:48 I wonder if it's due to a firewall. 12:52:00 DARN, forgot to run screen. :) 12:52:04 brb 12:52:06 --- quit: kc5tja ("leaving") 12:52:25 dig +trace doesn't work 12:52:41 --- join: kc5tja (~kc5tja@66-74-218-202.san.rr.com) joined #forth 12:52:41 --- mode: ChanServ set +o kc5tja 12:53:51 dig +trace didn't work (from home where it does resolve) just said: ;; connection timed out; no servers could be reached 12:54:22 Well, yeah, you have to do it outside of your own local network. 12:54:35 However, that error message you got is still indicative that something really wonky is going on with the name servers. 12:54:53 Have you checked to see if ns1.jwoof.net is up and running? 12:54:59 yeah. 12:55:04 direct queries work find 12:55:05 direct queries work fine 12:55:13 hmmm 12:55:20 eg: dnsq a jason.herkamire.com jwoof.com 12:55:52 I don't have dnsq 12:56:10 like dig 12:56:11 Are you on cable or DSL? 12:56:24 I just happen to have it's calling conventions memorized 12:56:28 cable 12:56:31 behind a router 12:56:45 I'm wondering if the cable provider is now blocking ports. 12:56:48 I have most things forewarded to my computer, but not everything 12:57:03 we have non-stop problems with that for our customers, where port 25 is blocked by a layer-4 switch inside their cable or DSL provider's network. 12:57:11 no, because T0paz got it, just now, and I can get it from my server (which is on long island NYC) 12:57:35 hmmm 12:57:47 yeah, some providers are total dicks about not letting communications through. 12:58:09 the only thing I've noticed blocked by my provider is connections going to my computer on the SMTP port 12:58:15 not sure how I noticed that... 12:58:40 they tell you not to run a webserver, but they don't even block port 80 12:58:46 ;; QUESTION SECTION: 12:58:46 ;ns1.jwoof.com. IN A 12:58:46 ;; AUTHORITY SECTION: 12:58:46 jwoof.com. 10 IN SOA ns1.jwoof.com. hostmaster.jwoof.com. 1087411217 16384 2048 1048576 2560 12:58:54 I'm just plain not seeing your nameservers. 12:59:05 And neither are the root name servers. 12:59:11 that's very bad 12:59:47 In fact, no whois record for jwoof.com -- did the domain expire? 13:00:00 I got a whois for it. it says expires april 2005 13:00:47 http://ws.arin.net/cgi-bin/whois.pl -- type in "jwoof.com", and you'll see that ARIN themselves don't see anything. 13:00:50 :) 13:01:02 --- join: TheBlueWizard (TheBlueWiz@modem-084.nyc-tc03a.FCC.NET) joined #forth 13:01:02 --- mode: ChanServ set +o TheBlueWizard 13:01:06 re TheBlueWizard 13:01:13 kc5tja hiya 13:01:19 Herkamire: You know, we charge our customers big bucks to do the diagnosis I'm providing. :D (j/k) 13:01:57 :) 13:02:21 Herkamire: My suggestion at this point is to mail-bom...er...talk to your registrar and have them re-instate your domain. :) 13:02:47 No whois match for herkamire.com either at arin.net. 13:02:53 I don't see why the dig +trace thing doesn't work 13:03:34 whois doesn't get all the registrars anymore 13:03:35 If it can't find your name server, it will either loop infinitely (in which case the search ends when the TTL of the request drops to zero), or it will just give up right away, saying, "no name servers found." 13:03:40 herkamire.com is registered with godaddy 13:04:12 Yes, but sometimes, godaddy and other registrars don't talk to their parent registrars about what is in their databases either. 13:04:26 right 13:04:28 ARIN.net, when it gives a successful whois, will tell you who the domain is actually registered with. 13:04:49 dig +trace says: ;; connection timed out; no servers could be reached 13:05:01 dig (without the +trace) works fine and prints the ip address etc 13:05:26 Is jwoof.com also registered with GoDaddy? 13:05:50 I was referred to whois.aitdomains.com; I'm looking it up there. 13:05:50 No Match for jwoof.com 13:05:51 [If E-mail address(es) were hidden on this page, you can click here to get the results with the E-mail address 13:06:03 I bet you it's all boiling down to jwoof.com 13:06:27 ok, now I'm starting to see problems 13:06:36 Hehe :D 13:06:54 I'm getting an initial whois for jwoof.com, but then it checks the refering server, and no match there 13:07:02 jwoof.com is from aitdomains.com 13:07:11 Our customers run into this problem all the time. They're constantly forgetting to pay their domain registration bills, OR, the registrar is constantly "testing" customers to see if domains are legit. 13:07:14 sounds like y'all learned a lot about DNS machinery :) 13:07:20 I was reluctant to transfer it because it has my precious name server registrations 13:07:40 TheBlueWizard: it's my job 13:07:41 * kc5tja nods 13:08:18 TheBlueWizard: I'm at work too, so that helps a bit. :) 13:09:42 * TheBlueWizard has realized that in order to effectively support the new "arbitrary" IP addressing that he (and likely kc5tja; he hopes they are on the same page), a DNS-like architecture is even more important...he read somewhere that "DNS is the worst naming scheme ever invented...except that all other naming schemes suck even worse :)", which he found it quite interesting on its own 13:10:10 * TheBlueWizard has a "surface" understanding of DNS...but nitty-gritty details, no 13:10:38 DNS is truely retarded, but I agree, X.500 and LDAP is far, far worse. 13:11:01 Although, LDAP/X.500 is more database savvy -- you can do searches that aren't necessarily hierarchially oriented. 13:11:04 It's more relational. 13:11:33 So, something like "jason.herkamire.com" might be something like "tld=com; domain=herkamire; machine=jason" in X.500 (well, they like to use shorter names, but you get the idea) 13:11:57 Interestingly enough, LDAP still manages to maintain data in a hierarchial format. 13:12:07 I'm on the phone with aitdomains now 13:12:08 But I've never gotten into the nitty gritty details (I tried once...failed). 13:13:20 this guy doesn't really seem to know much about computers 13:13:43 * TheBlueWizard nods...he knows a very little bit about LDAP...it has some nice ideas though, and tiny bits of such ideas should be, um, pilfered for the new network scheme :) 13:14:15 Hehe -- yeah. 13:14:22 * kc5tja fully agrees. 13:16:17 so looking up jason.herkamire.com it would ask for the nameservers for herkamire.com, then query them for the address of jason.herkamire.com 13:16:17 actually I am thinking about writing up a table like chart listing each networking technologies and their pros and cons, in order to help identify which ones should be looked at more closely 13:16:42 Right. 13:17:00 TheBlueWizard: Makes sense. 13:17:23 And you know, although I've said that I'm shying away from the ATM-model of networking, it does seem that there is strong merit to connection-oriented networking. 13:17:44 I mean, yes, packet-based networking is great and all, but as you get higher and higher in speed, it becomes increasingly hard to deal with. 13:17:52 of course it is a big effort, so some teamwork will be required 13:18:04 Even 10-gig-Ethernet strongly advocates connections via the extensive use of MPLS. 13:18:05 hmm 13:18:25 never heard of MPLS 13:18:38 Also, realistically speaking, except for radio and free-space optical, all network connections (even packet-type) involve point to point links of some kind. 13:18:57 TheBlueWizard: MPLS takes all the great ideas of ATM, and applies them to Ethernet frames. 13:19:15 So you can route/switch Ethernet packets very, very quickly, based on virtual circuit IDs. 13:19:28 Unlike ATM, though, you can nest "labels," (their name for VCIDs). 13:19:43 Which is kind of neat. I'm personally not sure why you'd want to do that, but I suppose they had a reason for it. 13:22:51 time to get a name server on a different domain 13:23:31 different registrar 13:23:42 Herkamire: :( 13:23:58 Are they just dicking you around or something? 13:24:02 What's their side of the story? 13:24:11 dunno 13:24:17 I didn't get anybody terribly technical 13:24:35 I got him to write down what the problem was and try to get it to somebody 13:24:56 now I have a use for my gmail account :) 13:25:03 Heheh :D 13:25:11 gave that to them in case herkamire isn't resolving properly 13:25:19 I got a big, blue bar on my gmail saying I have six invitations still to go. 13:25:38 before what? 13:25:41 * kc5tja is wondering if it's there because they're about to terminate the invitation program or not. I dunno. But all of my friends are already on gmail. :) 13:25:45 Herkamire: No idea. 13:25:56 I just never seen that bar there before, presented like it's some dire emergency. :) 13:26:05 seems to me everybody is going to have an account pretty soon 13:26:13 heh 13:26:27 clever marketing tactic 13:27:48 TheBlueWizard: The user interface of gmail rocks. 13:28:03 I really like it. The first real innovation in e-mail clients since the well-known 3-pane view. 13:28:10 But I digress. 13:28:26 I still would like my e-mail local, so I still keep all my important messages via traditional mail interfaces. 13:29:42 yeah, the gmail interface is very impressive 13:29:59 hmm 13:31:05 * kc5tja is reading up on garbage collection techniques again. 13:31:53 * kc5tja is thinking of making a simple, garbage collected heap implementation in GForth just to play with the concept. Maybe then I'll use that to make a simple e-mail client with it. 13:32:15 I wonder how fast I would exhaust a 1MB heap. :) 13:33:01 Oh, and pointers to objects would be kept in handles, so that I can access objects via @ @ and @ !, respectively. This way, I don't have to conservatively search the CPU's state or data/return stacks. 13:33:22 So it won't be 100% "automatic" memory management, but it will be automatic enough to enjoy playing with, I think. 13:34:59 sure 13:36:12 I do still find it interesting that copying and/or compacting collectors can still be extremely runtime efficient though. 13:36:43 * kc5tja will need to do more study on it, of course, but initial intuition suggests something like a mark-n-sweep garbage collector would actually be more efficient. 13:37:12 slava would likely know more about this than most others here, since he's implemented garbage collection for his Factor language. 13:37:28 yes...there exists some systems/languages with continuous gc, so that really make them go fast....a good example is Euphoria 13:37:54 Interestingly enough, Commodore 64 and 128 BASIC used compacting garbage collection. 13:38:10 The C64 BASIC was inefficient though, because it waited for the whole variable space to fill up before compacting. 13:38:25 This gave C64 BASIC the image of being incredibly slow for heavy data processing. 13:39:08 The C128 BASIC, however, had *incremental garbage collection*, which for the time, was utterly unheard of. C128 BASIC was the indisputed king of garbage reclamation speed too. Whole new implementation of the GC resulted in incredible performance gains. 13:39:13 I've encountered, and read about certain pathological case with gc on C-64...it can take hours to wait for gc to finish (I think it's published in a Compute! Gazette magazine) 13:39:18 grrrrr... I hate registrar issues 13:39:18 I think it did a variable or two of garbage every VSYNC interrupt or something. 13:39:50 TheBlueWizard: I don't know about hours. But even so, several minutes is not unheard of. 13:40:01 C128, to my knowledge, has *never* had to "stop" 13:40:03 'sopd'as 13:40:07 C128, to my knowledge, has *never* had to "stop" for GC to complete. 13:40:23 And then we get . . . Java. What a step backwards as far as GC is concerned. 13:40:34 But the BASIC implementation of GC could cheat though. 13:40:45 It stored every new variable instance on the heap. So if you say something like: 13:40:50 A$=A$+"WORLD" 13:41:03 you'd have two instances of A$ on the heap. One with just HELLO, and one containing HELLOWORLD. 13:41:04 it can happen on C-64...I've seen that happened on my C-64 after following some code shown in a magazine...I ended up turning it off :) 13:41:27 When it garbage collected, it'd just walk the list and if it found a newer instance of a variable later on, it'd collapse the heap. 13:41:33 (well, the C64 implementation did) 13:41:41 mmhmm 13:41:55 I'm sure the C128 had more efficient means of walking the variable list in the heap. 13:42:34 It probably could identify the front-most (e.g., the most recent) definition right up front, and therefore, could single-handedly collapse every other instance of the variable in one fell swoop. 13:43:33 It also helped out that the only structured variables that existed were arrays too -- no structures meant no pointers, and therefore, no need for tracing sub-structures. :) 13:43:41 I still have a hardcopy disassembly of AppleSoft Basic, and the gc code is one of perhaps two major areas yet to be decoded by me :) It would be reasonable to assume C-64 has essentially same algorithm 13:44:08 Not sure; never dealt with AppleSoft BASIC. 13:44:33 But, I know the author of EhBASIC, a modern BASIC implementation for 6502 and 65816 systems. I wonder what kind of garbage collection mechanism he uses there. 13:44:39 * kc5tja should ask him. 13:44:44 don't forget that you can have multidimensional arrays in Basic...that is a tiny bit of substructuring :) 13:44:53 kc5tja, factor's current gc won't scale with a large heap 13:45:03 kc5tja, its not incremental at all, and all live objects are copied on every collection 13:45:25 So it's not incremental, and it's not generational either. 13:45:38 nope:) 13:45:46 for what its worth, i never see gc pauses with java 13:46:37 slava: your computer's probably alittle faster than the C64 13:46:37 I have. 13:46:39 They're not fun. 13:47:03 We also see them on our servers here, and when it happens, it more often than not effectively takes the server right offline. 13:48:01 the nice thing about a copying collector is that running time depends on the number of live objects, not the amount of garbage. 13:52:33 on x86, does the icache have to flushed explicitly after generating code? 13:55:40 I think if the generated code has never been run before and the target memory area has never been JMPed into before, I don't think icache needs to be flushed. Best to consult an x86 doc 13:58:36 slava: Not explicitly, but it helps. 13:59:25 The reason is that any writes to memory that is located in the code cache WILL cause the code cache line to automatically flush. Likewise, attempting to execute code from memory currently in the d-cache WILL cause that d-cache line to flush as well. 13:59:36 This is why it is a big performance hit to run code and reference data in the same cache line. 13:59:55 yes, and why i'll be putting code in a separate heap ;) 14:00:13 FTS/Forth couldn't care less right now anyway. In practice, it is rarely an issue for me. 14:00:36 really? 14:00:48 how much of a performance hit is it? 14:00:53 When I say it isn't an issue for me, I mean that I don't particularly care about run-time performance right now. 14:00:58 ok 14:01:07 see i do, that's why i'm contemplating this compiler :) 14:01:10 I haven't measured, and I don't care. It doesn't affect the code I write. 14:02:56 Most of my code is so I/O bound that it really doesn't matter for me. Besides, there are ways even in FTS/Forth to make code that runs fast. 14:02:56 slava: then imitate kc5tja's methodologies...like writing up a test suite and run it, whatever....I'm sure kc5tja has done a lot of these 14:03:14 TheBlueWizard: Nothing for performance. I've never really benchmarked it. 14:03:28 * kc5tja might do benchmarking in the future once FTS/Forth for Linux is actually completed. 14:03:37 TheBlueWizard, i already have a 2000+ line test suite :) 14:04:00 oh? I thought you did, especially with testing the effect of using different registers on the threaded codes 14:04:08 Nope. 14:04:15 Oh, well, hold on. 14:04:17 That's different. 14:04:43 I did do tests involving different registers, because I wanted to see what instruction sequences ran the fastest for things like pushing and popping values off the data stack. 14:04:51 * TheBlueWizard has a rather good memory of that conversation 14:04:58 The data stack is inherently a different chunk of memory anyway, so it's not really relavent here. 14:05:10 It's always accessed at top speed. 14:05:22 (unless you're crazy enough to execute machine code pushed onto it) 14:09:51 slava, TheBlueWizard: can you reach http://ns1.jwoof.com/ ? 14:10:36 An error occurred while loading http://ns1.jwoof.com/: 14:10:36 Unknown host ns1.jwoof.com 14:10:45 how about http://jwoof.com/ 14:11:00 that loads 14:23:28 kc5tja: internic has my nameservers: http://reports.internic.net/cgi/whois?whois_nic=ns1.jwoof.com&type=nameserver 14:23:40 does that mean the root servers do? 14:25:09 If I do a dig +trace ns1.jwoof.com, M.GTLD-SERVERS.NET returns a response for that domain name. 14:25:17 However, without the +trace, I get nothing for a response. 14:25:38 Is 204.89.131.33 the correct IP address for jwoof.com? 14:25:49 yes 14:27:53 brb -- have to label a bunch of new servers 14:32:05 labeling servers... sounds like something I would do 14:32:46 at my last job I would put a sticker on the front of the computers with their name and IP address, in a vain hope that people wouldn't ask me anymore 14:32:49 didn't really work. 14:33:17