12:46:43 <cheater> hi!
12:47:46 <cheater> i want to proxy through to a happstack, however i will potentially have a lot of instances running and i don't want to use up thousands of ports. is it possible to have happstack talk to the server via a file?
12:49:49 <donri> cheater: i don't quite understand the question
12:50:12 <luite> donri: have happestack run on a unix socket instead of a port
12:50:25 <luite> with yesod you can do that if you supply your own socket
12:50:28 <cheater> i have a lot of happstack based services (gitit). say 1000. I want them to be accessible under, say, 1.domain.com through 1000.domain.com.
12:50:30 <luite> perhaps happstack has something similar
12:50:59 <cheater> i know you can use unix sockets in general, i need to figure out if happstack has this ability
12:51:14 <donri> i don't see how that's related to unix sockets, but you can run the server on a pre-initialized socket
12:51:27 <cheater> what is a "pre-initialized socket"?
12:52:05 <cheater> what i mean donri is that i tell happstack #1000 to use /var/happstack1000.sock instead of localhost:11000
12:52:33 <donri> something like, sock <- listenOn $ UnixSocket "/var/happstack1000.sock"; simpleHTTPWithSocket sock nullConf handlers
12:52:46 <donri> you'll need to import Network as well
12:52:52 <cheater> that sounds like what i'm looking for
12:53:06 <cheater> thanks a lot, donri
12:53:56 <donri> ah i see because you're using something like nginx in front of multiple happstack servers
12:54:08 <cheater> right yes
12:54:30 <cheater> the idea is to run gitit in my hg repos
12:54:40 <cheater> and then have a separate gitit instance available for every revision
12:54:50 <luite> o_O
12:55:00 <cheater> naturally those get loaded and shut down
12:55:09 <cheater> it's just so that the revs are available
12:55:25 <donri> doesn't gitit do that itself?
12:55:33 <cheater> yes and no
12:55:50 <donri> http://darcs.net/_history/FrontPage
12:55:55 <cheater> yes it allows you to load older revisions, no it does not let you see what the version of everything was for a specific revision
12:56:14 <donri> ah
12:56:17 <cheater> so you can't say "give me the wiki as it was on fc425dae"
12:56:58 <luite> hm, i'd rather patch gitit slightly than do this multiple instances thing
12:57:20 <cheater> i already have planned to supply access to separate gitit's via the branch name, so having the revision too isn't too big a difference
12:58:03 <cheater> right, that is also possible
12:58:12 <cheater> i'll toy with the possible ideas
12:58:13 <luite> no once you have the head pointer of a branch you have a revision and if oyu implemented revisions, you were already done
12:58:24 <luite> (that's assuming the git model, i'm not familior with hg)
12:58:45 <cheater> but even if i have that functionality in gitit, i still need sockets :)
12:58:54 <cheater> because there are 100s of repos here :)
12:59:41 <luite> right, i'm not saying that you don't, just that running gitit for each revision sounds a bit backwards
12:59:58 <cheater> yeah it does doesn't it
13:00:07 <donri> surely at least you'll run them in the same process though?
13:00:16 <cheater> not sure yet, i might
13:00:34 <cheater> i've seen the relevant doc page on Network.Gitit
13:01:02 <cheater> on hackage that is
13:01:19 <cheater> they show how to serve multiple gitits from a single happstack
13:02:28 <luite> does gitit use the command line tools for hg to retrieve the pages, like it does (did?) with git?
13:03:59 <cheater> i'm not sure
13:04:02 <cheater> i just installed it :p
13:04:05 <cheater> i'm guessing yes
13:05:54 <donri> looks like it
13:05:55 <donri> https://github.com/jgm/filestore/blob/master/Data/FileStore/Mercurial.hs
13:06:04 <cheater> here's how you hook up happstack to multiple gitits btw http://hackage.haskell.org/packages/archive/gitit/latest/doc/html/Network-Gitit.html
13:06:45 <donri> although looks like it can reuse hg processes https://github.com/jgm/filestore/blob/master/Data/FileStore/MercurialCommandServer.hs
15:20:24 <donri> stepcut: how's it going
15:20:36 <stepcut> good
15:22:46 <donri> :)
16:03:10 <donri> new boomerang release eh
16:04:56 <donri> nice new features!
16:10:01 <stepcut> no
16:10:06 <stepcut> just a bug fix
16:10:16 <stepcut> Texts.isComplete was missing the, isComplete _ = False, case
16:10:28 <stepcut> though.. that is possibly a temporary thing anyway..
16:10:46 <stepcut> I think we want to change how boomerang handles the idea of 'complete' parses
16:11:05 <stepcut> by default, a parse should only succeed if it consumed the entire list of path segments
16:11:14 <stepcut> but sometimes you actually want to allow leftovers
16:11:47 <stepcut> for example, if you have a constructor like, ThemeData, you might want to leave the remaining path segments alone and pass them on to serveDIrectory
16:12:04 <donri> (i know, i checked hdiff ;))
16:12:09 <stepcut> ah
16:12:17 <donri> you spawned some discussion in #haskell though
16:13:19 <donri> yea, serving static assets via web-routes is harder than it should be atm
16:13:55 <donri> even if you use serveFile instead, you still can't easily "get the rest of the path including slashes"
16:14:00 <stepcut> right
16:14:08 <stepcut> discovered that while implementing clckwrks
16:15:04 <stepcut> hmm, possible fire alarm, bbiab I hope
16:15:44 <stepcut> just a test.
16:15:45 <donri> o_O
16:16:09 <donri> please don't die in a fire stepcut!
16:16:22 <donri> or by drowning for that matter!
16:16:39 <stepcut> anyway, I am now writing about boomerang for a presentation I am giving on tuesday
16:18:08 <donri> cool, will there be slides/video published?
16:19:02 <stepcut> eventually
16:19:25 <stepcut> this is the first draft of my CUFP presentation
16:20:04 <stepcut> the slides will be useless on their own
16:20:15 <stepcut> but I will publish them with commentary in some form or another
16:22:05 <donri> nice
16:25:23 <stepcut> it needs to be easier to create PathInfo instances using boomerang though
16:29:21 <donri> agreed
16:29:42 <donri> preferably two-way
16:30:28 <stepcut> yes
16:56:43 <donri> argh so apparently my attempts with fontconfig worked... but not in firefox, which is where i was testing it.
16:57:19 <stepcut> :)
16:57:46 <donri> (i configured PT Sans to render with autohinting and everything else with normal hinting)
17:02:55 <stepcut> heh
17:03:27 <stepcut> i just left all the defaults alone
17:04:17 <donri> yea but osx has defaults that are half-crappy for all fonts instead of awesome for a few ones, and you're used to that and like it
17:04:38 <stepcut> yup
17:04:42 <donri> :)
17:04:53 <stepcut> but I have to design for the masses, not the elite :p
18:59:00 <mightybyte> stepcut: I just plugged acme-http for you. :)
18:59:15 <stepcut> sweet!
18:59:44 <mightybyte> It seemed to be appropriate for what Tekmo was looking for in his web server tutorial post.
19:00:03 <stepcut> :)
19:00:36 <mightybyte> Good stuff
19:01:00 <stepcut> I have something pretty close to acme-http built around pipes-core too.. when tekmo release the new pipes parser stuff I'll give it another go