[SR-Users] Is 't_newtran()' needed when using reliable TCP/TLS transport?

2018-04-11 Thread Cody Herzog
Hello. I have read that the primary use of 't_newtran()' is to handle SIP retransmissions. Given that, is there any reason to use it when operating over TCP/TLS? I recently changed a route to operate asynchronously through 'async_task_route()'. When I did, calls to 't_newtran()' started failin

Re: [SR-Users] Is 't_newtran()' needed when using reliable TCP/TLS transport?

2018-04-11 Thread Alex Balashov
Well, no, there are lots of things you need transaction state for. Failover, for example. If you tried gateway #1 and it timed out, so you want to try gateway #2, you have no way of doing that unless you have some way of knowing and interrogating the facts that: 1) gateway #1 was tried 2) a cer

Re: [SR-Users] Is 't_newtran()' needed when using reliable TCP/TLS transport?

2018-04-11 Thread Cody Herzog
Thanks for the quick reply, Alex. I will try to provide more context information. The route in question is related to presence. Our usage of t_newtran() is very similar to the following presence examples: https://github.com/kamailio/kamailio/blob/master/test/unit/presence.cfg https://lists.kama

Re: [SR-Users] Is 't_newtran()' needed when using reliable TCP/TLS transport?

2018-04-11 Thread Cody Herzog
Another thought... Perhaps 'async_task_route()' automatically creates a transaction, which is why I was having problems with the calls to 't_newtran()' and 't_release()' after the route had been changed to execute through 'async_task_route()'. If so, then maybe it's safe for my async presence r