00:31:55 <stepcut> :)
00:32:07 <stepcut> I wrote to niklas today about trying to get a new HSP out
00:32:30 <sm> yay
00:32:48 <sm> HSP deserves more love
00:34:29 <stepcut> yeah
21:08:55 <donri> stepcut: i'm reading http://okmij.org/ftp/Haskell/Iteratee/IterateeIO-talk-notes.pdf , seems relevant to hyperdrive and using pipes or not. have you read it?
21:10:02 <stepcut> yup, shortly after it was published
21:10:13 <stepcut> will re-read it after pipes 2.5 is out
21:10:50 <stepcut> greg was saying that doc (or a similar one) gave 4 reasons for using iteratees, but that in his opinion 2 of them were completely bogus
21:13:13 <donri> greg who?
21:14:09 <donri> http://ro-che.info/ccc/15.html
22:23:52 <donri> stepcut: hm re generic serialize, not sure it's such a big deal because e.g. lists will use a builtin instance, and how often do you make custom linked lists and not serialize them as lists? but, still, it's also worse than we previously thought because sums are serialized basically in binary but using a byte for each bit... so e.g. a third constructor would be something like, "\000\000\001"
22:24:49 <donri> i don't really know myself how to solve either issue
22:28:07 <donri> cereal itself seems to solve the latter issue with some crazy CPP and type level stack computations... and doesn't really have the former issue at all due to not having versions
23:13:29 <Lemmih> Hah, meteor is fun.
23:13:36 <Lemmih> semantique.tv
23:13:44 <Lemmih> meteor frontend, haskell backend.
23:15:08 <donri> and obvious bootstrap is obvious ;)
23:15:47 <donri> i have yet to see a bootstrap site that i couldn't tell was a bootstrap site. of course, if i ever do, i wouldn't know ...