Hi everyone,
Working with some actor system authors we realized we have a missing hook for cleaning up resources that are created on demand for a remote reference.
This was by design initially, however it turned out it’s not very good in practice and we need to have the ability to release resources held by the score system when the deinit of a remote actor reference runs.
This proposal introduces a new resignRemoteID in addition to the existing resignID (that applies only to locally assigned ids).
More details in:
I expect this to be rather boring for most folks, but input very welcome!
FYI @hassila et al, since you might be interested in this change for package-distributed-system maybe? I saw your resignID does fail on a remote being passed there, so that's another datapoint why we're not changing the existing resignID semantics but adding a separate one