2014-03-04 23:42:07 +0000 <carter> has a time out and runs it in a seperate haskell thread
2014-03-04 23:42:29 +0000 <hyperthunk> carter: ok
2014-03-04 23:42:32 +0000 <carter> http://hackage.haskell.org/package/base-4.6.0.1/docs/System-Timeout.html
2014-03-04 23:42:36 +0000 <hyperthunk> why do we need that?
2014-03-04 23:42:41 +0000 <hyperthunk> how does that solve the problem ?
2014-03-04 23:42:44 +0000 <carter> so youc an give a decent error instead of crash
2014-03-04 23:42:50 +0000 <hyperthunk> it's not the sender crashing that is the problem!
2014-03-04 23:42:54 +0000 <carter> well
2014-03-04 23:42:55 +0000 <hyperthunk> it is the receiver crashing
2014-03-04 23:42:57 +0000 <carter> ok
2014-03-04 23:42:58 +0000 <carter> well
2014-03-04 23:42:59 +0000 <hyperthunk> that's what should never happen
2014-03-04 23:43:09 +0000 <carter> but how'd it get there
2014-03-04 23:43:11 +0000 <carter> ok
2014-03-04 23:43:13 +0000 <carter> well
2014-03-04 23:43:14 +0000 <hyperthunk> semantically, it makes no sense for me to receive a bit of data from some (potentially remote) peer
2014-03-04 23:43:22 +0000 <hyperthunk> and then blow up because it contained 'undefined'
2014-03-04 23:43:25 +0000 <hyperthunk> on the other hand
2014-03-04 23:43:25 +0000 <carter> false
2014-03-04 23:43:28 +0000 <carter> computations do that :)
2014-03-04 23:43:40 +0000 <hyperthunk> it makes perfect sense for me to crash if I try to send something that's not fully evaluated
2014-03-04 23:43:49 +0000 <hyperthunk> carter: you're being very speculative there
2014-03-04 23:43:52 +0000 <carter> i know :)
2014-03-04 23:44:01 +0000 <hyperthunk> :)
2014-03-04 23:44:05 +0000 <carter> i'm just pushing to try to give alternative views
2014-03-04 23:44:10 +0000 <carter> even if they're wrong / bonkers
2014-03-04 23:44:26 +0000 <carter> :)
2014-03-04 23:44:29 +0000 <hyperthunk> I really don't think we need to boil the ocean over this. Really what we need is super fast serialisation and to force people (through kind words) to use NFData
2014-03-04 23:44:46 +0000 <hyperthunk> people who use NFData and d-p-platform have nothing to worry about
2014-03-04 23:44:56 +0000 <hyperthunk> especially if they stick to strict types for all messages
2014-03-04 23:45:07 +0000 <carter> hyperthunk: you forget, i like boiling oceans
2014-03-04 23:45:11 +0000 <carter> thats all i know how to do
2014-03-04 23:45:23 +0000 <carter> very very slowly boiling planets too
2014-03-04 23:45:30 +0000 <hyperthunk> like I said, the one's really suffering at the moment are the guys that are doing lots of real distributed stuff and have to pay the penalty for serialisation being so slow
2014-03-04 23:45:42 +0000 <carter> yes