This, I think, is a bit more natural to the OTP environment, since all of the code works with PIDs instead of our arbitrary session IDs. We should save a few calls to gproc and back this way, though performance-wise I'd estimate this is in the noise.
The session ID is still generated and can be looked up via ow_session:id/1, since it is useful when dealing with the wire format.
I also chipped away at reconnecting sessions a bit more. It's more complete but not function, I'd guess.
Change Overworld sessions to use process PIDs as the primary identifier, such that invocations like:
are now more like this:
This, I think, is a bit more natural to the OTP environment, since all of the code works with PIDs instead of our arbitrary session IDs. We should save a few calls to gproc and back this way, though performance-wise I'd estimate this is in the noise.
The session ID is still generated and can be looked up via
ow_session:id/1, since it is useful when dealing with the wire format.I also chipped away at reconnecting sessions a bit more. It's more complete but not function, I'd guess.