01:43 <SeanTAllen> emilbayes: there is a branch for the distributed pony work

01:44 <SeanTAllen> actually sylvanc merged that in. he's been doing it a bit at a time and merging in as we go. the first bit was the serialization work.

01:46 <SeanTAllen> on the distributed pony front related to sockets, files etc, what sylvan and i have discussed is being able to tag a resource as "machine specific" such that if your actor relies on a machine specific type like File, TCPConnection etc it can't be migrated. And user code could also take part in that system.

01:48 <SeanTAllen> some of the general ideas for distributed pony have already been implemented in our product we are building at Sendence. Sylvan has had a look and likes the general idea which makes sense because we model it after what distributed Pony might be. We spent a good amount of time talking it over with Sylvan.

03:08 <emilbayes> SeanTAllen: That's awesome, I'll check it out. Didn't even consider to check the branches of the repo. Pinning down resources seems like it would not make the system too magical. How about scheduling actors to minimise network traffic? Is that something the compiler / scheduler would be able to work out? How about nodes dropping off and coming back online?

03:08 <emilbayes> How would that be handled? Is it possible to mark work as optimistic/pessimistic?

13:03 <SeanTAllen> Nothing has been done for node join/leave yet emilbayes. Nor schedulig to minimise network traffic, I dont think that should be the responsibility of the runtime. What Sylvan and I have talked about doesn't hide the fact that you are making a remote call from you.

14:47 <SeanTAllen> hi all, I just opened an RFC for a small change to Map.upsert. Feel free to weigh in: https://github.com/ponylang/rfcs/pull/27