[Devel] [ openser-Bugs-1631420 ] another CANCEL bug

[Klaus Darilion]

Juha Heinanen wrote:
>  > If I recall correctly, theis problem was debated several times ago. It is
>  > more a logical issues than a bug and was labeled as "delayed CANCEL" - as
>  > the trace shows, it is about a transaction with a non-responsive branch,
>  > transaction wthat gets completed. After that a reply from the
>  > non-responsive branch arrives. The questions is what should you do?? If the
>  > transaction is still in memory, you can generate a delayed
>  > CANCEL.....otherwise the reply will not match any transaction and it will
>  > be stateless fwd....Any thoughts???
> 
> yes, if transaction state still exists, it makes sense to generate a
> delayed cancel.  
> 
> but if transaction does not exist anymore, i don't think it makes sense
> by a atateful proxy to statelessly forward any replies.  in other
> worlds, if in my proxy i always t_relay requests, i would like to be
> able to tell the proxy not to never statelessly forward any replies.

Me too. That's why I asked to check for transactions in reply route and 
drop final responses too: 
http://openser.org/pipermail/devel/2006-December/005001.html

regards
klaus

-- 
Klaus Darilion
nic.at