Benedikt Radtke

Not sure if multitargeting is worth the hassle, but we could include a sample impl in the documentation (some documentation would be great anyway, now that I think of it)....

[This](https://github.com/signal-csharp/libsignal-protocol-dotnet/blob/master/libsignal-protocol-dotnet/SessionCipher.cs#L168) decrypt variant should fit your needs: It calls the callback first, and only after the callback has completed the updated session is saved. If the process is killed or...

You are right, libsignal-service-java and dotnet **don't** use it. Since this should affect Signal-Android, maybe we should open an issue there and observe how OWS will fix it? This should™...

>Then the session data for later function calls (ie: sessionCipher.getSessionVersion) is not yet added to the session and the application will throw an exception with message "No session" when the...

What exactly do you mean with protocol changes and lack of replay? I just started up S-W, breaked in OnMessage (before the call to .Decrypt), killed the app, restarted it...

I am afraid that is working as intended - consumers get one chance to handle the message. If the OnMessage callback finishes with anything except a TaskCancelledException, we confirm the...

>I am not sure why if the process terminates after the decrypt call you would be unable to decrypt the replayed message? The only reason you can't decrypt is if...

>Here is a sample flow that would be bad: Client calls Decrypt with a callback. Decrypt starts process, calls callback with data (before counter is incremented) Client cannot save off...

Won't this deadlock if you call decrypt on a task with a synchronization context? Do we need asynchronity here so far down in the signal stack?

It only happened during the first start, so it was empty i think. It was at work, cannot check right now.