2014-03-04 23:21:53 +0000 <hyperthunk> what is the question?
2014-03-04 23:22:17 +0000 <carter> how do you syntactically ensure i can only send unserialized values on a shared heap channell
2014-03-04 23:22:27 +0000 <hyperthunk> carter: I see
2014-03-04 23:22:33 +0000 <hyperthunk> that is sorta useful
2014-03-04 23:22:35 +0000 <carter> yes
2014-03-04 23:22:47 +0000 <carter> this is a key key piece of my experimental design i hope to finish up
2014-03-04 23:22:51 +0000 <hyperthunk> but really, it pales in comparison IMHO to being able to serialise and send fast over the wire
2014-03-04 23:22:53 +0000 <carter> i want to *reason*
2014-03-04 23:22:56 +0000 <hyperthunk> which currently, we're very slow at
2014-03-04 23:22:57 +0000 <carter> about when that is so
2014-03-04 23:23:02 +0000 <hyperthunk> carter: sue
2014-03-04 23:23:04 +0000 <hyperthunk> sure*
2014-03-04 23:23:05 +0000 <carter> well
2014-03-04 23:23:10 +0000 <carter> user builder
2014-03-04 23:23:13 +0000 <carter> *use builder
2014-03-04 23:23:16 +0000 <hyperthunk> but right now, your cluster is going to suck because remote messaging isn't fast enough
2014-03-04 23:23:17 +0000 <carter> for anything you're serializing
2014-03-04 23:23:19 +0000 <carter> well
2014-03-04 23:23:20 +0000 <carter> thats ok
2014-03-04 23:23:23 +0000 <carter> i've plans for that too
2014-03-04 23:23:24 +0000 <carter> :)
2014-03-04 23:23:26 +0000 <hyperthunk> carter: great - how? wanna do a pull request?
2014-03-04 23:23:33 +0000 <hyperthunk> that'd be wicked, if it's faster
2014-03-04 23:23:33 +0000 <carter> once its done
2014-03-04 23:23:46 +0000 <hyperthunk> also, we've got a 200% binary overhead in the format we use
2014-03-04 23:23:47 +0000 <hyperthunk> urgh
2014-03-04 23:23:47 +0000 <carter> i'm stilling just writing down the plans
2014-03-04 23:24:10 +0000 <hyperthunk> carter: I can take a PR that makes things faster without any discussion, as long as the tests still pass.
2014-03-04 23:24:10 +0000 <carter> I will share things once they work :)
2014-03-04 23:24:16 +0000 <hyperthunk> carter: cool, that'd be good.
2014-03-04 23:24:32 +0000 <carter> one thing is i worry the current repos are over factorized
2014-03-04 23:24:33 +0000 <hyperthunk> the data families thing, I'm certainly keen to look at.
2014-03-04 23:24:36 +0000 <carter> yes
2014-03-04 23:24:50 +0000 <carter> i'm focused on getting my numericla haskell stuff out this month + income
2014-03-04 23:24:52 +0000 <hyperthunk> carter: d-p-platform is not, it's monolithic because I could not be bothered splitting it into a million bits