Johannes Rudolph
Johannes Rudolph
My fix in #30043 only worked around the `doWrap` issue but as already shown originally, there also seems to be a similar issue with `doUnwrap` that needs a similar treatment.
> My fix in #30043 only worked around the `doWrap` issue but as already shown originally, there also seems to be a similar issue with `doUnwrap` that needs a similar...
[2.6.19+37-4aeafc32-SNAPSHOT](https://oss.sonatype.org/content/repositories/snapshots/com/typesafe/akka/akka-stream_2.13/2.6.19+37-4aeafc32-SNAPSHOT/) should have the potential improvement from #31321 in case someone wants to try.
> I think of where we ended up on with this as a more generic operator which can do all that statefulMapConcat can by composing with mapConcat. I think is...
> once it has been out for a while and any eventual bugs has been found and squashed. Yes, no need to hurry with this one. :+1:
https://github.com/akka/akka/runs/6167388221?check_suite_focus=true#step:5:36778
It seems currently the main reason for dropped buffers is that there's a race condition between a shard becoming aware of a shard region being moved and propagating messages in...
I also like @johanandren's suggestion to use `StopShard` which seems less intrusive.
TLS 1.3 allows [0-RTT early client data](https://www.rfc-editor.org/rfc/rfc8446#section-2.3) which is not secured against replay attacks (because it is encrypted using a fixed PSK before DHE is finished). Auth data will only...
I can rule out a 0-RTT case, it seems to happen if the first data arrives in the same TCP packet as the `Finished` TLS handshake packet.