Async Push: Parsing message without a session about californium HOT 6 CLOSED

Scandium currently requires client code to synchronously establish the initial DTLS session with the server before sending arbitrary (sync or async) requests over the established session.
If you do not do so Scandium starts mulitple concurrent handshakes with the server (one for each async request) until one of the handshakes has finished.
We need to properly document this and maybe also prevent client code from submitting more requests until the session is established.

from californium.

Also see here for a description of what happens under the hood.

NB please do not comment on the referenced issue in californium.scandium because we have merged Scandium into the Californium repo.

from californium.

Is this still an issue? If so, I think one solution to fix it, would be adding and additional newDeferredMessageSender to the "ongoing handshake" instead of starting a new one.
Any opinions on this?

from californium.

@boacks I suppose this should works. (Even if I suspect some possible race condition)

from californium.

It seems to be no longer a issue of interest, so I would rather close than fixing it :-).

Even if this issue was not caused by coap retries during a ongoing handshake, such retries are from my experience mainly the cause for such problems. But in the meantime, such retries are not longer the case.

from californium.

Testing I found, that if "notifications" trigger a handshake (e.g Automatic session resumption is used), fast following notifications may be dropped.

So FMPOV, this issue should be fixed :-)

from californium.