00:21:16 <alpounet> donri, i *may* be flying to Sweden soon, you're living there right? if possible we should have a coffee or something, and complain about stepcut being late for the happstack 7 release
00:24:07 <DMcGill> right, next draft done
00:24:27 <DMcGill> I'd like to fill out the Source service and Local hackage sections a little more, but it's far too late now
00:24:41 <DMcGill> lectures tomorrow etc
00:25:15 <alpounet> DMcGill, seems good now, although it would be great if you made it clearer that you'd only (mostly) work on the builder and the dependency tracker
00:25:58 <alpounet> that's not obvious. we just understand that now from the fact that these are the only sections where you gave an outline of your plans for how much time will be spent
00:26:30 <DMcGill> I'll change that tomorrow afternoon. I should be around from 5 until 10 UK time and I really want to get it submitted it tomorrow
00:29:15 <donri> :D
00:30:37 <alpounet> DMcGill, sounds good, i'll be around from 5 to (at most) 7, UK time
00:32:21 <DMcGill> this is taking much longer than I expected, but it's shaping up really nicely :)
00:33:10 <alpounet> yeah, the proposal is waaaay better
00:43:01 <donri> alpounet: you use ghc7.4 too?
00:43:22 <donri> ACTION can't build scoutess on 7.0
00:43:24 <donri> the irony! :D
00:45:08 <alpounet> no, i have it somewhere but i don't USE it for my projects
00:47:06 <donri> ah, stepcut added the deps in cabal
00:48:25 <donri> darcs annotate sure has weird output ^_^
16:30:19 <stepkut> DMcGill: IMO, we are probably not going to be able to use hsenv, etc, off the shelf, but are going to need to copy the ideas to implement something that does what we need.. though that should only be around 1000 lines
16:36:35 <DMcGill> doesn't sound too bad then, particularly as we'll have the existing programs to see the methods from
16:36:43 <stepkut> exactly
16:36:45 <DMcGill> methods of*
16:36:52 <stepkut> there are 4 different apps that I know of to copy from
16:37:21 <stepkut> the problems I envision related to the fact that the apps are developer oriented and try to change the environment that the developer is working in
16:37:30 <stepkut> and they tend to provide poor isolation
16:37:43 <stepkut> so, you do something like, hsenv switch-to-env-for-some-app
16:37:51 <stepkut> and then ghc, etc, work in the new sandbox
16:38:01 <stepkut> but what if we have parallel builds going on on the same machine?
16:38:42 <stepkut> also, it can be tricky to get them to ignore the developers normal envirnoment -- we want to build in a completely isolated environment that is irrelevant to what the developers normal environment is
16:41:40 <DMcGill> I guess that's where the virtual env part comes in although I don't really know how they'd work
16:42:39 <DMcGill> presumably we can point cabal to specific package versions
16:43:19 <DMcGill> for the dependencies I mean, obviously you could do cabal install package-3.2.1 or whatever
16:56:18 <mekeor> so, Happstack is actually something like Apache+PHP, isn't it?  but with the difference that you have to compile it again after having changed it, right?
16:58:18 <stepkut> mekeor: in a sense
17:00:29 <mekeor> okay, nice, thanks :)
18:20:59 <DMcGill> Right, final draft!
18:21:00 <DMcGill> https://docs.google.com/document/d/1ulYTXJ8WlZz83-OUXMougEtlEjCLDAlSsfbXxg-Kyfs/edit
18:21:26 <DMcGill> I'll hand it in this afternoon unless anybody has objections
18:21:46 <DMcGill> Thanks for all your help and I look forward to hopefully working with you :)
19:14:10 <DMcGill> stepkut, the final draft is now up
19:14:22 <stepkut> DMcGill: cool!
19:14:36 <DMcGill> there was very little to do last time, this should be it
19:15:26 <stepkut> k
19:21:50 <DMcGill> although of course please let me know if you disagree!
19:31:51 <stepkut> donri: I see haskell-src-exts now supports DoIfThenElse.. is that do to your proding ?
19:35:34 <donri> :)
19:35:42 <donri> i suspect from reddit
19:35:49 <stepkut> ah
19:35:50 <donri> people just happened to be discussing the very same thing
19:36:09 <donri> now just need haskell-src-meta to support new -exts for jmacro
19:36:28 <stepkut> :)
19:36:42 <stepkut> time to send some emails! :)
19:36:45 <donri> which i filed an issue for
19:36:56 <stepkut> nice
19:36:57 <donri> though -meta seems rather active so
19:58:03 <stepkut> nice
20:04:03 <stepkut> DMcGill: looks good!
20:27:52 <stepkut> Lemmih: compared to multimaster replication, how hard would it be to make a 'remote backup client' that just logged events on a different server?
20:49:09 <donri> how is that different from replication
20:49:49 <donri> oh, you mean it'd be one-way
20:51:38 <donri> stepkut: you saw nibro's gsoc proposal?
20:52:31 <donri> all these nice ideas floating around and we have limited slots :(
20:56:29 <stepkut> donri: yup
20:57:01 <stepkut> donri: yeah, replication is two way and has sync issues. With my proposed system, the master should not block if the logger goes down..
20:57:35 <donri> "better than nothing"
20:58:03 <stepkut> if you have that.. and use the remote API... then you sort of have a master/client system I think
20:58:14 <donri> though somewhat missing the C of ACID, yea?
20:58:29 <stepkut> not sure
20:58:34 <stepkut> this is as far as I have thought
20:58:38 <stepkut> focused on the release mostly
20:58:42 <donri> oh i probably mean the D
20:59:15 <stepkut> that makes less sense..
20:59:41 <donri> well sounds like you can end up with different logs on master and backup?
20:59:56 <donri> which might still be better than not having any backup at all, but, yea
21:01:20 <donri> or perhaps I mean the "A" ;) in that the master and backup aren't atomic together
21:16:09 <Lemmih> stepkut: Much easier. Nearly trivial.
21:16:49 <Lemmih> stepkut: That's a good idea, actually.
21:18:12 <Lemmih> Yeah, hm. You don't get all the ACID guarantees.
21:18:59 <Lemmih> One could write a best effort state duplicator.
21:19:41 <Lemmih> Hm, don't like that at all, actually.
21:20:02 <Lemmih> Proper replication is the way to go.
21:43:20 <stepkut> :)
21:43:32 <stepkut> better get crackin on your GSoC submission :p
21:50:42 <donri> where do people get the idea that warp and snap is very fast
21:50:49 <donri> that is, they leave out happstack from that listing
21:52:10 <donri> i get 9k+ rqs from happstack on my dev machine. no idea if snap and warp would do more, but, happstack is not "slow" either way?
21:53:27 <stepkut> snap and happstack are around the same speed these days, last I checked
21:53:31 <stepkut> warp is 2-4 times faster
21:53:40 <stepkut> based on the single 'pong' benchmark
21:54:14 <donri> i watched some silk talk and he said they used both happstack and snap, "snap might be a little faster"
21:54:38 <donri> i mean, even if that is true, are they actually experiencing "happstack is too slow"?
21:54:42 <stepkut> no
21:54:50 <donri> what makes it even *worth noting* :P
21:55:16 <stepkut> one of the early snap promotition pieces was a benchmark showing how fast there were.. I think that is how it got started
21:55:21 <mightybyte> Nobody is saying that happstack is slow.
21:55:40 <mightybyte> At least nobody that I know of.
21:56:50 <mightybyte> And in my mind, comparison to warp is not really apples to apples.
21:57:16 <donri> mightybyte: i'm not saying people are saying happstack is slow. they're saying snap and warp are fast.
21:57:40 <donri> what the listener then *hears* is "the others are slow" :P
21:57:57 <stepkut> donri: fear not, Happstack 8 will be fast :)
21:58:04 <donri> and i just find that curious since they're all quite fast
21:58:30 <mightybyte> Well, that's a problem with the listener.
21:59:25 <mightybyte> Beware the ides of March!
22:37:25 <alpounet> DMcGill, proposal looks good!