2004/09/08
IRC [00:20] <brunom> Hey Dossy ... this bloody IRC program doesn't beep or anything ...
IRC [00:20] <brunom> Anyways, leaving as well ... bbl
IRC [01:23] *** zoro2 joined the chat.
IRC [02:41] *** tekbasse joined the chat.
IRC [02:46] *** zoro2 parted the chat.
IRC [03:55] *** frankie joined the chat.
IRC [04:25] *** zoro2 joined the chat.
IRC [06:41] <zoro2> ok. if anyone's alive
IRC [06:41] <zoro2> how do I port Ns_TclGetConn(NULL) to 4.0?
IRC [06:54] <zoro2> ok found it
IRC [07:19] *** frankie parted the chat.
IRC [07:25] *** zoro2_ joined the chat.
IRC [07:26] <zoro2_> hello again
IRC [07:26] <zoro2_> anyone alive? please :)
IRC [07:34] *** zoro2 parted the chat.
IRC [07:44] <Dossy> Ns_TclGetConn() ?
IRC [07:48] *** frankie joined the chat.
IRC [07:51] <zoro2_> ok nevermind as well
IRC [07:51] <zoro2_> but why were custom ADP parsers made obsolete in 4.0?
IRC [07:51] <zoro2_> this is going to make my life a PITA
IRC [07:52] <zoro2_> btw the good thing is that nsdqe now works on 4.0
IRC [08:15] <Dossy> heh
IRC [09:14] <zoro2_> I really need to learn AOLserver 4 - where can I read on encoding changes since 3.4
IRC [09:14] <zoro2_> ?
IRC [09:33] *** brunom parted the chat.
IRC [10:31] <Dossy> zoro: there really isn't a good guide - should write one for the wiki I guess ...
IRC [10:57] <zoro2_> yes
IRC [10:57] <zoro2_> I have a problem with utf-8
IRC [10:58] <zoro2_> set ns/parameters' OutputCharset and URLCharset to utf-8
IRC [10:58] <zoro2_> have a file in utf-8
IRC [10:58] <zoro2_> and it is sent back as iso8859-1
IRC [10:58] <zoro2_> with Content-Type: text/html; charset=iso-8859-1
IRC [10:58] <zoro2_> with Content-Type: text/html; charset=iso-8859-1
IRC [10:58] <zoro2_> and encoded wrong
IRC [10:59] <zoro2_> and ns/encodings' .dqe to utf-8 as well
IRC [11:00] <zoro2_> ok done it
IRC [11:01] *** zoro2_ parted the chat.
IRC [11:10] <martinh> hey. can I fconfigure the conn i'm writing on?
IRC [11:39] <Dossy> martinh: don't think so - although that'd be sweet.
IRC [11:41] *** zoro2 joined the chat.
IRC [11:54] *** rubick joined the chat.
IRC [12:09] *** zoro2 parted the chat.
IRC [12:10] <Dossy> man. meetup.com is falling apar.
IRC [12:10] <Dossy> er apart
IRC [12:19] *** frankie parted the chat.
IRC [13:14] *** GizmoBeastLives joined the chat.
IRC [13:14] <Dossy> hey jamie
IRC [13:15] <GizmoBeastLives> hi Dossy, thanks for looking into the nslog error,
IRC [13:15] <Dossy> Sure thing. Alex keeps asking me about the new auth module too :)
IRC [13:15] <GizmoBeastLives> I'll try to get you a better test, this is with 4.1.0a and no streaming in sight
IRC [13:15] <GizmoBeastLives> That doesn't surprise me :-)
IRC [13:15] <Dossy> try a fresh checkout/build of 4.1.0a maybe?
IRC [13:16] <GizmoBeastLives> Maybe, I certainly have lots of patches in the core, but I really don't think any are related,
IRC [13:17] <GizmoBeastLives> I can step through the code and watch this happening just as I described -
IRC [13:17] <Dossy> I'm sure - as I said, I *think* I fixed it a while ago.
IRC [13:17] <GizmoBeastLives> only thing I'm not certain of is what is causing the WriteConn to fail,
IRC [13:18] <GizmoBeastLives> what did you do to "immediately disconnect the socket"?
IRC [13:18] <Dossy> what do you have on line 2051 in nsd/driver.c?
IRC [13:18] <Dossy> I ^C'ed wget before the response, as well as when testing with telnet, ^]q to close telnet.
IRC [13:18] <GizmoBeastLives> connPtr->nContentSent = 0;
IRC [13:18] <Dossy> line 2121?
IRC [13:18] <GizmoBeastLives> connPtr->nContentSent = 0;
IRC [13:19] <Dossy> OK, well - that's what I originally thought might fix the problem. Oh well. :)
IRC [13:19] *** tekbasse parted the chat.
IRC [13:19] <GizmoBeastLives> cvs diff -> no changes to driver.c
IRC [13:20] <Dossy> nsd/return.c Ns_ConnQueueHeaders decrementing the nContentSent is a bad idea.
IRC [13:20] <Dossy> should probably have a nContentQueued or something
IRC [13:20] <Dossy> and incr that
IRC [13:20] <Dossy> and whent content is sent, incr nContentSent and decr nContentQueued
IRC [13:20] <GizmoBeastLives> yes, I think that's probably a better solution
IRC [13:21] <Dossy> Should eliminate this issue entirely ...
IRC [13:21] <Dossy> Or really, connPtr->obuf.length is the nContentQueued
IRC [13:22] <Dossy> It seems that headers aren't counted as sent content -- that's why nContentSent is decr by the header len
IRC [13:22] <Dossy> IS that the "desired" behavior?
IRC [13:22] <Dossy> I guess for "content" sent, headers are not content ...
IRC [13:22] <GizmoBeastLives> I think it is
IRC [13:23] <Dossy> but in the access log, it's *bytes* sent, which IMHO *should* include headers as they're bytes sent.
IRC [13:24] <Dossy> >> The last entry indicates the size of the object returned to the client, not including the response headers. If no content was returned to the client, this value will be "-". To log "0" for no content, use %B instead.
IRC [13:24] <Dossy> Aha, *not* including the response headers. Oh.
IRC [13:24] <Dossy> Good to know ...
IRC [13:24] <GizmoBeastLives> right- otherwise you couldn't compare filesize on disk with log, etc.
IRC [13:25] <GizmoBeastLives> Dossy, did you try with any longer sleep periods than 5 seconds?
IRC [13:26] <GizmoBeastLives> I tried your script with ns_sleep 60 and I get the problem, but I don't with ns_sleep 5
IRC [13:26] <Dossy> Interesting --
IRC [13:26] <Dossy> I'll give it a try.
IRC [13:28] <GizmoBeastLives> I don't see any odd timeouts configured in my config file
IRC [13:29] <Dossy> Hmm ..
IRC [13:29] <Dossy> what's default timeout? maybe 30sec?
IRC [13:29] <GizmoBeastLives> yeah, I think there is a default timeout of 30s for both send and receive maybe
IRC [13:30] <Dossy> Interesting.
IRC [13:30] <GizmoBeastLives> let me know if you can reproduce this, that would help conserve my sanity :-)
IRC [13:30] <Dossy> you're using sample-config.tcl?
IRC [13:30] <GizmoBeastLives> no, I'm using a real config file but nothing too odd in there
IRC [13:30] <Dossy> by the way - what do you think of the new "Configuration Options" section I added to the nscgi doc page: http://cvs.sourceforge.net/viewcvs.py/*checkout*/aolserver/aolserver/nscgi/nscgi.html?rev=HEAD
IRC [13:31] <Dossy> 10.6.113.2 - - [08/Sep/2004:13:30:45 -0400] "GET /nslog.adp HTTP/1.0" 200 224 "" "Wget/1.9.1" 60.001537
IRC [13:32] <Dossy> 60.001537 sec, bytes 224
IRC [13:32] <Dossy> that was with wget, trying with telnet now
IRC [13:32] <GizmoBeastLives> nscgi docs are nice
IRC [13:33] <Dossy> the text isn't pretty (needs a stylesheet, probably) but the content is now up-to-date with the code, and I tried to explain the config options
IRC [13:33] <Dossy> Going to try and do this for all the standard/core modules.
IRC [13:47] *** frankie joined the chat.
IRC [13:48] <Dossy> By the way, testing with telnet yielded same result in log.
IRC [13:48] <GizmoBeastLives> ok, I'll keep investigating
IRC [13:48] <Dossy> try reproducing with sample-config.tcl
IRC [13:48] <Dossy> and the test page
IRC [13:49] <Dossy> i'm upping the sleep to ns_sleep 120
IRC [13:52] <Dossy> 120sec, no diff
IRC [13:53] <GizmoBeastLives> hmm, maybe platform or client matters
IRC [13:58] <Dossy> Ns_SetIGet() -- what does it do if there's multiple values for the key? Only return the first/last?
IRC [13:58] <Dossy> maybe -- what platform are you testing on? win32?
IRC [13:58] <GizmoBeastLives> yes, win32
IRC [13:59] <GizmoBeastLives> I believe iget returns the first if there are multiple
IRC [14:41] <Dossy> OK -- just updated nslog doc
IRC [14:41] <Dossy> http://cvs.sourceforge.net/viewcvs.py/*checkout*/aolserver/aolserver/nslog/nslog.html?rev=HEAD
IRC [14:41] <Dossy> oh - crap - won't show up on anoncvs web for a few hours :(
IRC [14:44] *** cnk joined the chat.
IRC [14:49] <Dossy> hi cnk :)
IRC [14:54] <cnk> hello
IRC [14:57] <Dossy> man, I really want to fix bug #972994 ... but it could break LOTS of stuff potentially :(
IRC [14:57] <Dossy> cnk, do you know csari in person? or just coincidentally from the same area?
IRC [14:58] <cnk> No idea who csari is. Why do you ask.
IRC [15:10] <GizmoBeastLives> Dossy I think you just posted comments on the wrong bug
IRC [15:16] <jhavard> dossy: that's the plus sign bug
IRC [15:17] *** frankie parted the chat.
IRC [15:22] <Dossy> I thought csari is also a caltech'er
IRC [15:23] <Dossy> gizmo: huh?
IRC [15:23] <Dossy> wrong bug?
IRC [15:23] <Dossy> yes - the plus sign bug - I want to fix that badly
IRC [15:23] <Dossy> but I see potential for breaking backwards compatibility ...
IRC [15:26] <Dossy> And, any input on bug http://aolserver.com/sf/bug/935907 ? the issue of redirecting when the URL refers to a directory but doesn't end with "/"
IRC [15:27] <Dossy> and bug http://aolserver.com/sf/bug/999089 -- does anyone else have Netscape 4.06 installed? :)
IRC [15:30] <cnk> Dossy: that would be interesting but I don't know anyone with that handle - and don't see it in a quick directory search.
IRC [15:32] <GizmoBeastLives> Dossy: your note about ns_log was supposed to go on #815850 I think, instead it went on #972994
IRC [15:32] <Dossy> cnk - never mind :)
IRC [15:33] <Dossy> gizmo- oh, ack :(
IRC [15:33] <Dossy> I blame it on SF bugs! : P
IRC [15:38] <Dossy> Yes - you're right -- thanks for letting me know.
IRC [16:03] <jhavard> You know dossy, if you didn't generate so many bugs, you wouldn't have to fix so many.
IRC [16:04] <jhavard> Who cares that 98.31% of the bugs are from before your time
IRC [16:04] <jhavard> it's all your fault.
IRC [16:05] <Dossy> yes, it is!
IRC [16:05] <Dossy> actually, i was around back since 1999 :)
IRC [16:05] <Dossy> but they're not my fault, but i should have fixed 'em all years ago :P
IRC [16:05] <Dossy> but alas, my paying job interfered with my aolserver dev time :P
IRC [16:05] <Dossy> now, my paying job IS my aolserver dev time. :)
IRC [16:06] *** iharding joined the chat.
IRC [16:07] <jhavard> so, how about that module I whipped up for uuid?
IRC [16:10] <iharding> The logging robot seems to have fallen down.
IRC [16:10] <iharding> qdf
IRC [16:15] <Dossy> ugh.
IRC [16:15] *** leff parted the chat.
IRC [16:15] <Dossy> no, the logging bot is ok --
IRC [16:15] <Dossy> I saw your mention of the uuid module --
IRC [16:16] <Dossy> rather than having a stand-alone nsuuid module, I'd like it to be a part of the core and just ship with aolserver
IRC [16:16] <jhavard> of course, then you're making a dependency on e2fs-progs
IRC [16:16] <Dossy> you have any issues with that?
IRC [16:17] <jhavard> that's the main issue.
IRC [16:17] <Dossy> I think we should clone Apache's "contributor agreement form"
IRC [16:17] <Dossy> e2fs-progs? EEK
IRC [16:17] <Dossy> OMG
IRC [16:17] <jhavard> yeah
IRC [16:17] <Dossy> i'd rather depend on libuuid
IRC [16:17] <jhavard> well, that's where libuuid comes from
IRC [16:17] <Dossy> there's no reason that we can't implement uuid from scratch, it's trivial :)
IRC [16:17] <Dossy> ok - never mind :)
IRC [16:17] <jhavard> true
IRC [16:17] <Dossy> i'm gonna work up some uuid routines.
IRC [16:18] <jhavard> I just needed to fill an immediate need I had and wanted to see how easy it was to work up a module.
IRC [16:18] <jhavard> It's not even beyond my C skills.
IRC [16:22] <Dossy> no - it's quite easy actually.
IRC [16:22] <Dossy> Tcl was designed from the start with embeddability and extendability in C in mind.
IRC [16:22] <Dossy> Other languages make it "possible" to extend them in C. Tcl was intended to be simple and lightweight and easy to extend in C.
IRC [16:23] <jhavard> I gathered that rather quickly.
IRC [16:23] <Dossy> If you want, link to your nsuuid on the wiki on the Modules page
IRC [16:23] <jhavard> 90% of the code in the nssha1 module was actually for sha1
IRC [16:23] <jhavard> 8% was actually comments
IRC [16:24] <jhavard> and 2% was the glue
IRC [16:24] <Dossy> yup.
IRC [16:27] <jhavard> http://www.opengroup.org/dce/info/draft-leach-uuids-guids-01.txt
IRC [16:29] <Dossy> in my todo bin I have an RFE for ADP caching
IRC [16:29] <Dossy> introducing a new [ns_adp_cache] command.
IRC [16:29] <jhavard> ooh
IRC [16:30] <Dossy> I'm also working up the spec for a PROPER session management system ... along with the "nssic" module - SIC meanign Server Inter-Connect protocol ...
IRC [16:30] <Dossy> so you could use that to do clustered session management
IRC [16:30] <Dossy> dont' know how far I'll get on all this -- it's my 2005 list of things to do
IRC [16:32] <jhavard> ooh
IRC [16:34] <jhavard> I guess I could work on the uuid module
IRC [16:34] <jhavard> and make it not lameness dependent
IRC [16:35] <Dossy> yeah, start by implementing it in pure tcl
IRC [16:35] <Dossy> then optimize into C :)
IRC [16:37] <jhavard> hrm, random number generator for tcl?
IRC [16:40] <jhavard> nevermind
IRC [16:40] <jhavard> found it
IRC [16:50] <jhavard> http://www.opengroup.org/dce/mall/dcecp-hcell.htm
IRC [16:50] <jhavard> oh great
IRC [16:50] <jhavard> dce and tcl
IRC [16:52] <Dossy> expr rand()
IRC [16:52] <jhavard> I already found it
IRC [16:53] <Dossy> that url docesn't really give you a UUID though
IRC [16:53] <jhavard> right
IRC [16:53] <jhavard> just stating that they seemed to have been big on tcl
IRC [17:12] <Dossy> crap. soda i spilled on my 'w' key today is starting to stick ... grr
IRC [17:15] <jhavard> The keyboard I use at work is like that on the numpad
IRC [17:16] <jhavard> my predecessor managed to spill a couple of gallons of coke on the keypad
IRC [17:16] <jhavard> I've done a marvelous job cleaning it out, but the 8 key still sticks
IRC [17:17] <jhavard> bah, my first attempt at a tcl uuid proc can only do 12000/second
IRC [17:20] <Dossy> how many do you need? :)
IRC [17:20] <jhavard> A few hundred a day.
IRC [17:20] <jhavard> oh
IRC [17:20] <jhavard> I can do better, though.
IRC [17:26] <jhavard> okay, I'm up to 16000/second
IRC [17:27] *** ian joined the chat.
IRC [17:27] <ian> can't read "tcl_version": no such variable
IRC [17:27] <ian> set tcl_library [file join /usr/pkg/lib tcl${tcl_version}
IRC [17:27] <ian> This used to work....
IRC [17:31] <Dossy> eek - 52 usec/iter
IRC [17:32] <Dossy> that's still 19,230/sec though :)
IRC [17:32] <Dossy> ian: heh -- weird
IRC [17:32] <Dossy> what version of tcl? 8.5?
IRC [17:32] <ian> 8.4, but this is my first real attempt at aolserver 4
IRC [17:33] <ian> I have something borked....can't read "tcl_platform(platform)": no such variable
IRC [17:33] <ian> while executing
IRC [17:33] <ian> "string equal $tcl_platform(platform) "windows""
IRC [17:33] <ian> (file "/usr/pkg/lib/tcl8.4/init.tcl" line 1)
IRC [17:35] <jhavard> hrm, i guess it is different since I'm using mostly random numbers instead of time and mac address
IRC [17:35] <Dossy> $::tcl_platform(platform) --
IRC [17:36] <Dossy> jhavard: i'm using random too - implementing version 4
IRC [17:39] <ian> Crud! That's wierd. I can source ...init.tcl and package require works
IRC [17:39] <jhavard> the interesting thing about the spec is that it doesn't say "this is the only way", merely that it's the suggested way
IRC [17:40] <ian> But you are right, the line it is choking on is missing the :: although
IRC [17:40] <ian> they are there in all other references.
IRC [17:40] <jhavard> well, not even that. It's more like, "this is how it's implemented on this, and in theory, it should be unique if all of these are followed"
IRC [17:40] <jhavard> but that's only for version 1
IRC [17:44] <ian> Ha ha! Using a 3.4.2 init script on 4.0.7. Doh!! Going home now.
IRC [17:45] *** iharding parted the chat.
IRC [17:45] *** ian parted the chat.
IRC [17:45] <Dossy> i'm down to 18,867/sec
IRC [17:46] <Dossy> http://panoptic.com/wiki/aolserver/nsuuid
IRC [17:46] <Dossy> it's far from optimal, but it's really straightforward
IRC [17:46] <Dossy> shit, and it's incorrect
IRC [17:51] <jhavard> http://www.opengroup.org/dce/info/draft-leach-uuids-guids-01.txt includes a reference implementation in c
IRC [17:52] *** GizmoBeastLives parted the chat.
IRC [18:26] *** zoro2 joined the chat.
IRC [18:27] <zoro2> hello again
IRC [18:41] <Dossy> yeah, it's easy to implement in C
IRC [18:41] <Dossy> the point is implementing it in pure tcl as a challenge :)
IRC [18:43] *** cnk parted the chat.
IRC [18:51] *** bartt joined the chat.
IRC [18:52] <rubick> Dossy: are you going to announce 4.08 on www.aolserver.com ?
IRC [18:55] *** zoro2 parted the chat.
IRC [19:00] <Dossy> oh - i did in SF news
IRC [19:00] <Dossy> apparently SF's project cron daemon still isn't running
IRC [19:00] <Dossy> let me update b hand
IRC [19:01] * rubick downloads 4.08 to install.
IRC [19:01] <rubick> ls
IRC [19:01] <Dossy> ok - ran it by hand - it's in the Latest News now
IRC [19:01] <Dossy> ok, i'm heading out for the night.
IRC [19:01] <rubick> Thanks.
IRC [20:13] *** rubick parted the chat.
IRC [20:34] *** brunom joined the chat.
IRC [20:44] <Dossy> Evening ...
IRC [20:45] <bartt> evening Dossy
IRC [21:02] *** brunom parted the chat.
AIM [21:05] *** sn3 webdesigner joined the chat.
AIM [21:05] <sn3 webdesigner> Hey does anyone own a Mac?
AIM [21:05] *** sn3 webdesigner joined the chat.
IRC [21:29] *** brunom joined the chat.
IRC [22:25] *** bartt parted the chat.