2005/04/01
IRC [00:09] *** tektubby parted the chat.
IRC [01:48] *** holy_cow joined the chat.
IRC [02:03] *** holycow parted the chat.
IRC [02:34] *** holy_cow parted the chat.
IRC [02:35] *** frankie joined the chat.
IRC [02:38] *** holycow joined the chat.
IRC [04:06] *** holycow parted the chat.
IRC [04:18] *** tektubby joined the chat.
IRC [06:39] *** tektubby parted the chat.
IRC [08:59] *** martinh parted the chat.
IRC [09:02] *** martinh joined the chat.
IRC [09:39] *** frankie parted the chat.
IRC [14:41] *** bartt joined the chat.
IRC [15:25] *** cnk joined the chat.
IRC [15:32] <Dossy> hey, cnk :-)
IRC [15:33] <cnk> hi
IRC [15:34] <cnk> getting back to our load testing stuff
IRC [15:34] <cnk> and about to scream in frustration about the state of the aolserver docs
IRC [15:35] <cnk> it seems to me that some of the tuning parameters might be overspecified - but that is probably because I am not understanding what they stand for
IRC [15:35] <cnk> and the last decent set of docs about the parameters file was for 2.3
IRC [15:36] <cnk> is there anything more detailed than the annotated configuration file?
IRC [15:37] <cnk> I am trying to figure out how one can set connsperthread, maxthreads, and maxconnections and not run into a situation where you can ask for a mathematical impossibility
IRC [15:38] <Dossy> more detailed? there's the source code ... it doesn't lie ...
IRC [15:39] <Dossy> connsperthread right now is dangerous as it seems to leak memory.
IRC [15:39] <cnk> is the way out of this that maxconnections and maxthreads are maximums at any given point in time but connsperthread is just a counter?
IRC [15:39] <Dossy> yes
IRC [15:39] <Dossy> connsperthread - after this many conns are serviced, the thread gets destroyed and a new one created
IRC [15:40] <cnk> I really need to get onto a more recent version of AOLserver so it is easier for me to read source and recompile at will
IRC [15:40] <cnk> and the "destroy and recreate threads" is itself leaking memory?
IRC [15:40] <cnk> bummer
IRC [15:41] <Dossy> it's similar in concept to apache's MaxRequestsPerChild directive
IRC [15:41] <Dossy> yeah. it's a SF bug
IRC [15:41] <cnk> I thought the point of connsperthread was to destroy threads that potentially had leaks
IRC [15:42] <Dossy> maxconns ~= apache's MaxClients
IRC [15:42] <Dossy> oh, wait no --
IRC [15:42] <Dossy> that's MaxThreads
IRC [15:42] <Dossy> aolserver maxthreads ~= apache MaxClients
IRC [15:42] <Dossy> aolserver connsperthread ~= apache MaxRequestsPerChild
IRC [15:42] <siddfinch> Well any suggestions to improve the docs, forward them to me
IRC [15:43] <siddfinch> BTW Dossy, want to give me a quick Intro to AOLserver for the LJ article?
IRC [15:43] <Dossy> or, aolserver maxthreads = apache ThreadsPerChild
IRC [15:44] <siddfinch> Dossy that is the way I understood it maxtheads = threads/child
IRC [15:44] <Dossy> intro to AOLserver? uh - not sure what you mean
IRC [15:44] <cnk> My main suggestion for the problem I am having now is to document what the parameters are, i.e. the implication of changing one of them, what the default is, and what a reasonable range is. The last is kind of hard but perhaps could be handled by "this set will serve XXX static pages per second on yyyy hardware"
IRC [15:45] <cnk> I would be glad to contribute my config and load test data - if I ever get a config that is decent
IRC [15:45] <Dossy> cnk: heh
IRC [15:45] <Dossy> cnk: the problem is you're load testing OpenACS, not bare AOLserver, right?
IRC [15:45] <Dossy> did you see the Litespeed benchmarks?
IRC [15:46] <siddfinch> For the first article I was going to describe the history of AOLserver and wanted to say something like ``Current AOLserver head huncho Dossy the Man wants you to know this about AOLserver ... ''
IRC [15:46] <cnk> I am testing on original ACS - but can run some numbers on bare AOLserver too. No I didn't see those benchmarks, is there a link from the AOLserver site?
IRC [15:46] <siddfinch> (or something along those lines)
IRC [15:48] <siddfinch> I just figured I'ld include a quote or two from you in the introduction ... let people understand that AOLserver is truly open source and people are involved.
IRC [15:48] <cnk> any news on the Tcl connection handling explanation that was mentioned on the mailing list? (I am behind on mail so did not chime in "yes please" there)
IRC [15:49] <Dossy> siddfinch: Oh. Today's not a good day for sound bites. :-)
IRC [15:49] <Dossy> cnk: huh? are you talking about Jim's additional commenting
IRC [15:50] <siddfinch> Dossy: well think about it and send me an email when you can .. I'll be outlining the web articles this weekend and the print article sometime next week
IRC [15:50] <cnk> there was the discussion about what could and could not be covered by additional commentting
IRC [15:50] <cnk> I think the conclusion was that some archetecture decisions were to complex for comments
IRC [15:51] <Dossy> siddfinch: intro to aolserver is really what's already up on aolserver.com --
IRC [15:51] <cnk> was the conclusion that Jim would try to comment the code rather than write up separate docs?
IRC [15:51] <cnk> If so, guess it is time to do a CVS checkout and start reading
IRC [15:51] <Dossy> yes, and he did write comments, and committed them, and asked for feedback and has gotten ZERO.
IRC [15:52] <Dossy> it's not exactly positive feedback and encouragement for Jim to continue commenting the code with his very limited time.
IRC [15:52] <Dossy> I still think that the code is remarkably clear and concise, and if you don't understand the design from the code, it's not a comments-are-necessary issue.
IRC [15:52] <cnk> OK then I know what my afternoon plans have to be - install RHEL 4 and then checkout and try to compile AOLserver
IRC [15:52] <Dossy> It's more like a "you must be this tall to ride this ride" issue.
IRC [15:52] <Dossy> ooh, heh
IRC [15:53] <cnk> yes - and for those of us who are short (or at least short on time) it would be great to have admin docs that let us try optimizing our config without reading source.
IRC [15:55] <tekbasse> and the admin docs could be useful for summarizing features for marketing
IRC [15:55] <cnk> very
IRC [15:57] <Dossy> absolutely.
IRC [15:57] <Dossy> that's why I said that what would REALLY be valuable are high-level powerpoint slide type documents
IRC [15:57] <Dossy> 3-4 slides and a bunch of bullet points that give a high-level overview of how the stuff works
IRC [15:58] <Dossy> then, with some better annotated config docs, people could tweak and tune etc.
IRC [15:58] <tekbasse> with links to code (and comments) for the code readers =)
IRC [15:58] <cnk> I will probably want more detail than that - but if you would write the high-level overview, then some of us could take a stab at filling in the outline in a more detailed version
IRC [15:59] <Dossy> ah, if you read code, you grep.
IRC [15:59] <Dossy> sure
IRC [16:00] <tekbasse> but grepping doesn't work well if you're trying to explain the features and someone throws you a technical question...
IRC [16:00] <tekbasse> while in a board meeting etc
IRC [16:01] *** frankie joined the chat.
IRC [16:02] <tekbasse> references are a sign of good documentation writing practices..
IRC [16:02] <tekbasse> bbl
IRC [16:18] <Dossy> yeah, we need better docs. I need to write some.
IRC [16:18] <Dossy> I just /hate/ doc writing.
IRC [16:26] <martinh> all programmers hate doc writing.
IRC [16:34] <siddfinch> everybody hates writing docs .. unless they have mental problems that don't make them programmers
IRC [16:36] <Dossy> not really.
IRC [16:36] <Dossy> Jim's good at it.
IRC [16:37] <tekbasse> I like writing docs, just don't have time to do it because I'm too busy writing code 'cause that's more productive than trying to convince others of the usefulness of writing some code
IRC [16:39] * tekbasse plans to contribute to doc writing as soon as he finishes the current project/coding objectives.
IRC [16:51] <martinh> being good at it is different than liking it.
IRC [16:52] <Dossy> That's true.
IRC [16:52] <tekbasse> he he that's very true. That's why I try to be responsive to feedback
IRC [16:53] * tekbasse goes for a walk
IRC [17:40] *** frankie parted the chat.
IRC [17:40] <Dossy> okay, cutting and pasting LARGE amounts of data from MSIE to Excel doesn't work very well. :P
IRC [18:05] *** anlater joined the chat.
IRC [18:39] *** bartt parted the chat.
IRC [19:13] *** holycow joined the chat.
IRC [19:43] *** holycow parted the chat.
IRC [19:47] *** win joined the chat.
IRC [19:54] *** win parted the chat.
IRC [19:55] *** win joined the chat.
IRC [20:18] <Guilmon> alindeman: in face, can you just give that host a week kline, its an open proxy
IRC [20:18] <Dossy> can chanserv enforce bans?
IRC [20:19] <alindeman> Yah
IRC [20:19] <alindeman> /cs autorem #aolserver add <ban>
IRC [20:19] <alindeman> <ban> being a mask
IRC [20:19] <Dossy> aha.
IRC [20:21] <Dossy> ok added to autorem list -- thanks guys.
IRC [20:21] *** Guilmon parted the chat.
IRC [20:24] <martinh> i'm surprised there's spam bots on freenode. . .i mean, pissing off a bunch of smart tech savy geeks with little else to do seems counterproductive.
IRC [20:24] <alindeman> Every so often we get a few people who aren't familiar with our philosophies or area of focus
IRC [20:25] <alindeman> They just see "Oh ... here's a network with a couple tens of thousands of users"
IRC [20:27] <Dossy> martinh: I think the fact that we log the chat to the web has a lot to do with it.
IRC [20:27] <Dossy> Since our chat logs turn URLs into anchor tags in the HTML logs.
IRC [20:28] <Dossy> PageRank whores will come here and spam URLs for juice.
IRC [20:28] <Dossy> I thought about doing the whole rel="nofollow" to the links, but I actually like giving juice to URLs which deserve it.
IRC [20:28] <Dossy> brb.
IRC [21:17] *** win joined the chat.
IRC [21:17] *** win parted the chat.
IRC [23:10] <Dossy> Quiet night.