dedbox
2017-11-30 19:10:25

Ok, so I’ve moved onto messengers while the codec situation unfolds.


dedbox
2017-11-30 19:10:49

Messengers are easy to model directly from the roadmap.


dedbox
2017-11-30 19:23:24

How to prevent interleaving of concurrent reads and writes, though?


dedbox
2017-11-30 19:26:53

Some options off the top of my head: transaction start/stop primitives, or make messenger read/write calls “atomic.”


notjack
2017-11-30 20:26:58

@dedbox 1) planning on responding in more detail later, but about codecs: yes this absolutely sounds right, or at the least it’s the right foundation / direction


notjack
2017-11-30 20:34:50

@dedbox 2) messenger sends and receives will be atomic and not concurrent to a single messenger but I think there’s tricks to let you build multiple messengers that interleave over a lower-level messenger. Also see the “atomic box” implementation in the disposable package and the paper “Kill Safe Synchronization Abstractions”, with the former being a building block abstracting over the latter; that’s how I’m imagining messenger atomicity will be implemented