2004/07/13
IRC [00:07] *** jhavard parted the chat.
IRC [00:07] *** rcrit parted the chat.
IRC [00:07] *** jhavard joined the chat.
IRC [03:06] *** thecodemill|aol parted the chat.
IRC [03:55] *** frankie joined the chat.
IRC [04:19] *** Nizz joined the chat.
AIM [05:02] *** POLearyUK joined the chat.
AIM [05:05] *** usteinacker68 joined the chat.
AIM [05:06] *** usteinacker68 joined the chat.
IRC [05:28] *** Nizz parted the chat.
AIM [08:39] *** WebbWrestler joined the chat.
AIM [08:40] *** WebbWrestler joined the chat.
IRC [08:44] <Dossy> heh
AIM [08:44] <POLearyUK> morning
IRC [08:54] <Dossy> howdy pat.
AIM [08:54] <POLearyUK> howdy Dos
AIM [08:54] <POLearyUK> been quiet this morning, not sure that many people on this side of atlantic here
AIM [08:55] <POLearyUK> hey think it would be worth while doing a geographical survey of aolsevers users? serves no purpose, just curiosity
IRC [08:57] <Dossy> sure
IRC [08:57] <Dossy> on the wiki :)
AIM [08:57] <POLearyUK> :-)
IRC [08:58] <Dossy> Create yourself a home page. :)
IRC [08:58] <Dossy> we can create geographical categories, if people like.
IRC [08:58] <Dossy> wiki backreferences == insta-indexing.
AIM [09:00] <POLearyUK> we might tie it up with TCL conferences as well-
IRC [09:11] <Dossy> ?
IRC [09:57] *** leff joined the chat.
IRC [10:41] *** nate joined the chat.
AIM [10:42] <POLearyUK> hey Nate
IRC [10:44] <nate> hey Dossy?
IRC [10:44] <nate> oh hey POLearyUK
AIM [10:44] <POLearyUK> :-)
IRC [10:50] *** Nizz joined the chat.
IRC [10:52] <frodoroot> how many ppl are on the AIM bridge?
AIM [10:53] <POLearyUK> just me in the aim chat room right now
AIM [11:00] *** bart teeuwisse joined the chat.
AIM [11:00] *** bart teeuwisse joined the chat.
AIM [11:00] *** bart teeuwisse joined the chat.
AIM [11:01] *** bart teeuwisse joined the chat.
IRC [11:01] <Dossy> heh
AIM [11:02] *** DossyNJ joined the chat.
AIM [11:34] <DossyNJ> pat - saw your email
AIM [11:35] <POLearyUK> yep
IRC [11:37] *** frankie parted the chat.
AIM [11:39] <DossyNJ> Too many options can be a bad thing.
AIM [11:39] <DossyNJ> Especially if an option incurs a per-request cost.
AIM [11:40] <POLearyUK> true, but the design of if ! host && default_host, can't be that much can it?
AIM [11:41] <DossyNJ> for a default host? no --
AIM [11:42] <DossyNJ> however, if you're NOT doing virtual servers, there's no need for a default host:
AIM [11:42] <DossyNJ> or at least, I don't think there should be
IRC [11:42] <Dossy> nope, there isn't.
AIM [11:42] <POLearyUK> true- shouldn't be
IRC [11:42] <Dossy> it's only IF you have software virtual serving configured
IRC [11:42] <Dossy> supporting a default host IFF it's an HTTP/1.0 request AND no virtual matches - that's cheap performance-wise
AIM [11:43] <POLearyUK> yep; that's what apache guys do
IRC [11:43] <Dossy> with HTTP/1.1, Host: is mandatory.
IRC [11:43] <Dossy> if the supplied Host: doesn't match any known virtual, what shoudl we do: default host, or send an error?
IRC [11:43] <Dossy> lets look at the RFC
AIM [11:44] <POLearyUK> hmmm- have it somewhere here
IRC [11:44] <jcollins> don't know what the rfc says but it'd be nice if it did default houstst
IRC [11:44] <jcollins> host
AIM [11:45] <POLearyUK> - 4xx: Client Error - The request contains bad syntax or cannot be fulfilled
IRC [11:47] <Dossy> http://www.w3.org/Protocols/rfc2616/rfc2616-sec5.html#sec5.2
IRC [11:47] <Dossy> 3. If the host as determined by rule 1 or 2 is not a valid host on the server, the response MUST be a 400 (Bad Request) error message.
AIM [11:47] <POLearyUK> although, that's asuming that the host isn't running another interface with a down daemon though
IRC [11:47] <Dossy> the RFC is pretty clear about the behavior.
IRC [11:47] <Dossy> So, why deviate from it?
AIM [11:48] <POLearyUK> 3. If the host as determined by rule 1 or 2 is not a valid host on the server, the response MUST be a 400 (Bad Request) error message.
IRC [11:48] <jcollins> Can I do wildcards in virtual host setup in AOLserver 4.x?
IRC [11:48] <jcollins> So *.tclsource.org always defaults to the tclsource.org server?
AIM [11:50] <POLearyUK> should we refine that to say only if host matchs ip's or dns that daemon is bound to?
IRC [11:50] <Dossy> pat: huh?
IRC [11:50] <Dossy> jcollins: no, wildcards do not work. that's the discussion I want to start in the email I sent to the AOLSERVER list today.
IRC [11:50] <Dossy> jcollins: are you subscribed to the list?
IRC [11:51] <jcollins> yeah
IRC [11:51] <jcollins> i generally read it in batches
IRC [11:53] <jcollins> couldn't you bypass globbing if its not defined?
IRC [11:54] <jcollins> so take what you say in the email a step further
IRC [11:54] <jcollins> if that is easy
IRC [11:55] <jcollins> i really think apache defaults to something instead of erroring
IRC [11:56] <jcollins> maybe it's just our setup here is broken
AIM [11:58] <POLearyUK> hmm have just tried on apache 1.3.9
AIM [11:58] <POLearyUK> bash-2.05b# telnet acme.office.alallan.com 80GET / HTTP/1.1Host: bob.office.alallan.comHTTP/1.1 200 OKDate: Tue, 13 Jul 2004 15:44:13 GMTServer: Apache/1.3.9 (Unix) PHP/4.0.5 Resin/1.2.7Last-Modified: Tue, 27 Aug 2002 14:18:09 GMTETag: "560c-678-3d6b8a21"Accept-Ranges: bytesContent-Length: 1656Content-Type: text/html
AIM [11:59] <POLearyUK> gave 200, and default index
IRC [12:13] *** cd34 joined the chat.
IRC [12:14] *** cd34 parted the chat.
AIM [12:15] <DossyNJ> did it?
AIM [12:15] <POLearyUK> yep
AIM [12:15] <DossyNJ> OK, then an nsconfig parameter "strict" by default off.
AIM [12:15] <DossyNJ> If you turn it on, then if no match, you'll get a 400 Bad Request.
AIM [12:16] <DossyNJ> Otherwise, the default. Sound reasonable?
AIM [12:16] <POLearyUK> sounds like a plan
IRC [12:43] <jcollins> read jim's email
IRC [12:53] *** bartt joined the chat.
IRC [12:57] *** thecodemill|aol joined the chat.
IRC [12:57] * thecodemill|aol is logging
IRC [12:58] <frodoroot> the quick brown fox jumped over the lazy dog
IRC [13:01] <martinh> very nimble fox. . .
IRC [13:04] <Dossy> Hmm.
IRC [13:09] <frodoroot> very lazy dog. . .
AIM [13:20] *** POLearyUK joined the chat.
AIM [13:25] *** RSeeger00 joined the chat.
IRC [13:27] <jhavard> "the quick brown fox..." contains every letter except "s"
AIM [13:28] <DossyNJ> true.
IRC [13:28] <frodoroot> the quick brown fox jumped over the lazy dogs
AIM [13:28] <DossyNJ> unless you typo.
IRC [13:28] <jhavard> there we go. now we have every single letter. perfect.
AIM [13:36] <DossyNJ> grrr
IRC [13:37] <frodoroot> :)
IRC [13:37] <frodoroot> do foxes growl?
IRC [13:43] <jhavard> don't be mean to the poor little foxes.
IRC [13:43] <jhavard> unless they taste good.
IRC [13:52] *** cd34 joined the chat.
IRC [13:53] *** cd34 parted the chat.
IRC [13:53] *** cd34 joined the chat.
IRC [13:57] <Dossy> OK, so who is cd34?
IRC [13:57] <Dossy> :)
IRC [13:59] <cd34> me
IRC [13:59] <cd34> sorry, fat fingers on the mouse button
IRC [14:02] <cd34> I tried to close that AIM message that got sent on join, and hit the wrong X on the window :) three times...
IRC [14:02] <Dossy> hah
IRC [14:02] <Dossy> yeah, i'm torn about the on-join spamming.
IRC [14:02] <Dossy> but, it's all for the newbies. THINK OF THE NEWBIES!
IRC [14:02] <cd34> think of the people that can't figure out xchat haha
IRC [14:13] <Dossy> argh, pat disappeared.
IRC [14:37] <jhavard> hrm, if I use a namespace, will any variables in the namespace persist throughout the life of the thread, unless it is set again.
IRC [14:46] <jcollins> yes
IRC [14:47] <jhavard> well damn. that complicates things.
AIM [14:57] <RSeeger00> Is that a bug, or an intended thing?
IRC [14:58] <jhavard> a bug.
IRC [14:58] <jhavard> although easily worked around.
IRC [15:00] <jhavard> Just need to reset the array used to store template variables after rendering the template.
IRC [15:03] <jhavard> ooh, need to be sure to handle a null list in the list-of-sets iterator.
AIM [15:07] <DossyNJ> my branes hurt
AIM [15:07] <DossyNJ> need another smoke.
IRC [15:07] <jhavard> to make your lungs hurt?
IRC [15:07] <jcollins> haha
AIM [15:08] <DossyNJ> workign on the host:-header default thing
AIM [15:08] <DossyNJ> there's many BAD ways to hack it in
AIM [15:08] <DossyNJ> but the most elegant way, is not easy.
AIM [15:08] <DossyNJ> without adding an element to one of the data structures.
AIM [15:08] <DossyNJ> ugh
IRC [15:10] * cd34 wonders if Dossy knows he's talking to a bot haha
IRC [15:10] <jhavard> which would break countless other things?
IRC [15:11] <cd34> or is that the AIM gateway?
IRC [15:14] <jhavard> that's the aim gateway
AIM [15:15] *** El Havard joined the chat.
AIM [15:15] <El Havard> now, it could get quite strange if he was talking to himself.
IRC [15:15] <jhavard> elhavard: yeah, I know what you mean.
AIM [15:15] *** El Havard joined the chat.
AIM [15:15] <DossyNJ> :)
IRC [15:20] <jhavard> dossy: from the tcl2k presentation, it looks to me that nv's are just a central server pushing data to other servers for use as nsv's, is this correct?
AIM [15:23] <DossyNJ> basically.
IRC [15:35] *** Nizz parted the chat.
IRC [15:40] *** bartt parted the chat.
IRC [16:13] *** cd34 parted the chat.
IRC [16:23] <Dossy> ungh
IRC [16:24] <frodoroot> I know I know it's "hung"
AIM [16:24] <DossyNJ> heh
IRC [16:39] <Dossy> two cvs up's at the same time. i think sourceforge might break. :P
IRC [17:08] *** cd34 joined the chat.
AIM [17:14] <DossyNJ> fixin bugs a plenty today
IRC [17:25] <Dossy> so, what versions of aolserver is everyone running?
IRC [17:25] <frodoroot> 3.x
IRC [17:25] <frodoroot> and 4.0
IRC [17:26] <frodoroot> did you mean y'all or everyone?
AIM [17:27] <DossyNJ> y'all :)
IRC [17:27] <cd34> 7x 4.0.5, one running virtual with 6 openacs installations
AIM [17:27] <DossyNJ> cd34: so, everything 4.x. cool.
AIM [17:27] <DossyNJ> frodoroot - what's stopping the 3.x from being upgraded to 4.x?
IRC [17:27] <cd34> and the debian package at that :)
AIM [17:28] <DossyNJ> cd34: no way! that's cool.
IRC [17:28] <cd34> cvs and I don't get along, so, if its not packaged, and I can't get one of my guys to package it, I don't run it :)
AIM [17:28] *** GizmoBeastLives joined the chat.
AIM [17:28] <DossyNJ> I see.
IRC [17:29] *** bartt joined the chat.
AIM [17:29] <DossyNJ> Would you be one of the folks interested in nightly tarballs?
IRC [17:29] <cd34> nah, I like my interrupted sleep too much
AIM [17:30] <GizmoBeastLives> I think nightly tarballs would be great if it is easy to do - the first step to nightly builds, which are even more useful
AIM [17:34] <DossyNJ> True.
AIM [17:34] <DossyNJ> I'm keen on doing nightly builds, for sure.
IRC [17:37] <cd34> I think they are good for development, but I don't do enough development to be a test case...
AIM [17:38] <DossyNJ> True.
AIM [17:38] <DossyNJ> See, that's what gets me - nightly builds really don't benefit many people.
AIM [17:38] <DossyNJ> However, I think binary releases quickly following official releases is important.
AIM [17:38] <DossyNJ> That certainly will benefit many, who aren't ready or able to build their own binaries.
IRC [17:49] <frodoroot> DossyNJ: I am stopping 3.x from being upgraded to 4.x
AIM [17:50] <DossyNJ> Why?!
IRC [17:50] <frodoroot> I just have to do it
IRC [17:50] <frodoroot> nsopenssl was crashing it when I tried once
IRC [17:50] <frodoroot> now I have to get the new nsopenssl going first
IRC [17:50] <frodoroot> and then try cutting over
IRC [17:50] <frodoroot> and then if that doesn't work try again
AIM [18:05] <DossyNJ> nod
AIM [18:05] <DossyNJ> I personally tested (and commited changes into) nsopenssl 3.0beta20 --
AIM [18:06] <DossyNJ> worked well at AOL for us.
IRC [18:26] <frodoroot> ok well that's nex ton the list
IRC [18:26] <frodoroot> for Saturday
AIM [18:36] <DossyNJ> cool.
IRC [18:42] * leff is away: home
IRC [18:45] <frodoroot> is there an elegant way to have a proc timeout after a certain period and having it kill the programs it execd?
AIM [19:02] *** RSeeger00 joined the chat.
IRC [19:03] <bartt> frodoroot: How about the after command?
AIM [19:03] <DossyNJ> mm, yeah.
IRC [19:09] <frodoroot> what about after? isn't it the same as sleep?
AIM [19:12] <DossyNJ> nope
AIM [19:13] <DossyNJ> ghetto alarm:
IRC [19:13] <frodoroot> somehow i thought that was coming
AIM [19:13] <DossyNJ> oh - you want to cancel stuff? that's not gonna happen ...
IRC [19:13] <frodoroot> yeah want to cancel
AIM [19:13] <DossyNJ> currently no way to cancel a thread :)
IRC [19:13] <frodoroot> or at worst at least return early from the proc
AIM [19:13] <DossyNJ> you can return early, though
IRC [19:13] <frodoroot> it also execs commands on the machine from the thread
IRC [19:14] <frodoroot> I could still kill those right?
AIM [19:14] <DossyNJ> after idle {some long running command}
AIM [19:14] <DossyNJ> after $timeout {salarm 1}
AIM [19:14] <DossyNJ> oh oops
AIM [19:14] <DossyNJ> wrong
AIM [19:14] <DossyNJ> try that again:
IRC [19:14] <frodoroot> what's salarm
AIM [19:14] <DossyNJ> after idle {long running script; set ::alarm 0}
AIM [19:14] <DossyNJ> after $timeout {salarm 1}
AIM [19:14] <DossyNJ> vwait ::alarm
AIM [19:15] <DossyNJ> if {$::alarm} { puts "took too long. time out. } else { our command finished. yay. }
IRC [19:15] <frodoroot> tcl in aolserver has vwait?
AIM [19:16] <DossyNJ> sure
IRC [19:16] <frodoroot> sure?
AIM [19:16] <DossyNJ> i wouldn't recomend using vwait in aolserver, but you're already using exec, so who cares
IRC [19:16] <frodoroot> what's wrong with exec
AIM [19:16] <DossyNJ> forks the nsd.
IRC [19:16] <frodoroot> it's the life blood of our system
AIM [19:17] <DossyNJ> yuck
IRC [19:17] <frodoroot> :|
AIM [19:17] <DossyNJ> then use apache.
AIM [19:17] <DossyNJ> it's already multi-process.
IRC [19:17] <frodoroot> ha
AIM [19:17] <DossyNJ> with pre-forked children
IRC [19:17] <frodoroot> apache funny
AIM [19:17] <DossyNJ> I'm serious.
IRC [19:17] <frodoroot> I'm sure
AIM [19:17] <DossyNJ> AOLserver is not for you. :)
IRC [19:17] <frodoroot> it's a little late to be making fundamental changes like that
AIM [19:18] <DossyNJ> *shrug*
AIM [19:18] <DossyNJ> must not be high traffic sites, then ...
IRC [19:18] <frodoroot> thanks a lot now I won't be able to sleep tonight
IRC [19:18] <frodoroot> :)
AIM [19:19] <DossyNJ> :P
AIM [19:19] <DossyNJ> if you can live with constant fork/exec of your aolserver, that says a lot :)
IRC [19:19] <frodoroot> you have an audience now
AIM [19:20] <DossyNJ> who?
IRC [19:20] <frodoroot> so what exactly is wrong with execing a lot?
IRC [19:20] <frodoroot> three people here at work
AIM [19:20] <DossyNJ> As I said earlier, it forks the nsd. Forking a pthread'ed app is ugly because you bring across all the threads into the forked child, for one.
AIM [19:20] <DossyNJ> Forking is bad because your new process inherits the memory footprint of the original nsd process. Then, you go overlaying that with the new executable when you exec()
AIM [19:21] <DossyNJ> Linux has copy-on-write and hopefully it's smart enough not to copy the nsd mem just to get it overwritten by the new executable. If it's not, well, you incur the penalty of the unnecessary memcpy() followed by another memcpy() overwriting it.
AIM [19:22] <DossyNJ> All the advantages of AOLserver go out the window when you fork/exec, essentially.
AIM [19:24] <DossyNJ> There's some keen improvements I'm looking to make to the nscgi module to make it perform better, but that's low on my priority list.
AIM [19:24] <DossyNJ> Because right now, nscgi forks the nsd. Ugh.
AIM [19:25] <DossyNJ> Still there?
IRC [19:25] <frodoroot> yeah
IRC [19:25] <frodoroot> reading
IRC [19:25] <frodoroot> we're all discussing
IRC [19:27] <frodoroot> so what makes apache better at execs?
AIM [19:28] <DossyNJ> it starts a master (parent) process that's very thin/light
AIM [19:28] <DossyNJ> which then forks off many child processes which are also very light. which in turn exec() whatever the request requires.
AIM [19:29] <DossyNJ> also, it keeps a pool of pre-forked child processes, and can re-use them
AIM [19:29] <DossyNJ> so the overhead of the fork() itself is absorbed that way.
IRC [19:31] <frodoroot> any interesting
IRC [19:31] <frodoroot> I mean
IRC [19:31] <frodoroot> interesting
AIM [19:32] <DossyNJ> yeah.
AIM [19:32] <DossyNJ> However, if you implement your applications in pure Tcl/ADP, performance can generally crush any CGI-based solution in Apache.
AIM [19:33] <DossyNJ> Simply because thread creation is so much faster than process creation, and evaluating a Tcl script is faster than running a standalone program.
IRC [19:34] <frodoroot> how does aolserver compare with apache performance overall?
IRC [19:38] <frodoroot> well?
IRC [19:40] <frodoroot> I'll save it for the list
IRC [19:40] <frodoroot> chao
IRC [20:01] *** cd34 parted the chat.
AIM [20:02] <DossyNJ> sorry - was doing something eles
AIM [20:02] <DossyNJ> er else.
IRC [21:07] <Dossy> So, who's still awake? :)
IRC [21:08] * jhavard raises hand
IRC [21:08] <jhavard> although I'm not working on the blog module
AIM [21:15] *** DossyNJ joined the chat.
IRC [21:15] <Dossy> heh
IRC [21:15] <Dossy> i'm about to leave for the night
IRC [21:15] <Dossy> stomped a few bugs
IRC [21:18] <Dossy> bye everyone.
AIM [21:54] *** GizmoBeastLives joined the chat.