Drew

I will suggest in earnest that the CommunityToolkit's APIs be considered for parts of this; - In the case of managed buffers being allocated internally (e.x. remoting), the `MemoryOwner` Implementation...

> So I've been looking into doing this using those APIs here: [Aaronontheweb/SystemMemorySerializer#2](https://github.com/Aaronontheweb/SystemMemorySerializer/pull/2) > > I have a few reasons for thinking that this won't be feasible so long as...

> > > @to11mtm @Aaronontheweb thanks for the analysis in #5230 . Just to confirm do we expect this to improve performance even in the case of the following approach?...

@Aaronontheweb some notes: ### Regarding overhead of sharded actors/etc: - One benchmark worth considering would be a 'forwarded ask' benchmark in a local scenario. i.e. - (Ask) to Actor (A)...

> For what it's worth, I do have a 100% complete gRPC Akka.Remote implementation lying around that will fallback to working over HTTP 1.1, I think, but I wouldn't recommend...

@Aaronontheweb I can confirm this is also happening on a project I'm dealing with. I can say that when this happened on our system, -sometimes- switching back to the Non-batching...

I've managed to make a Repro but it definitely requires a bit of cleanup and it's messy. What I'm seeing is a pattern where under heavy writes a command within...

Some general-ish notes: - A return of `IMemoryOwner` rather than `ReadOnlyMemory` from `ToBinary()` would let consumers of the API control final disposal and/or reuse of the segment to maximize usage,...

Should we consider going away from `ArraySegment` for internal storage and use `Memory` instead? On one hand it would be a bit of a yak shave, on the other hand...

> > I don't think we will get any improvement switching over to Memory, even microsoft ReadOnlySequence still uses ArraySegment underneath. > > being able to work with new native...