Re: [reSIProcate] CANCEL and TransactionState
This isn't a bad idea. However, are there considerations that are
different for user agents from proxies? In the past, we've taken the
approach of putting nothing that is UA-specific or proxy-specific in
the transaction layer.
Doing the CANCEL processing in repro isn't too bad. The CANCEL
processing in dum is pretty involved though. I'm not sure how this
would end up refactoring.
Jason
On 7/16/07, Byron Campen <bcampen@xxxxxxxxxxxx> wrote:
I have been thinking about how the stack (and TransactionState in
particular) treat CANCEL from the TU. Right now, sending a CANCEL to
the stack is the only way the TU can express its desire to shed
INVITE transaction state, but the stack will only accept a CANCEL if
the transaction has received a response. This basically means that
the TU has to keep state around denoting whether it wants to let go
of a given INVITE transaction, whether it is allowed to send a CANCEL
yet, and whether it has already sent a CANCEL (replicating portions
of the transaction-state-machine). Might it be convenient to keep
this information down in TransactionState instead, since anything
that sends an INVITE will ultimately be forced to keep this state if
it wants to be protected from a deliberately non-responsive UAS?
Thoughts?
Best regards,
Byron Campen
_______________________________________________
resiprocate-devel mailing list
resiprocate-devel@xxxxxxxxxxxxxxxxxxxx
https://list.resiprocate.org/mailman/listinfo/resiprocate-devel