15 Jun
2011
15 Jun
'11
3:32 p.m.
2011/6/9 Iñaki Baz Castillo <ibc(a)aliax.net>et>:
2011/6/9 Alex Hermann <alex(a)speakup.nl>nl>:
The problem is fixed now, it was an error related to the NAPTR service
flag (sip-router expected it to be "s" case-sensitive, while it should
also allow "S").
The way I read rfc2915, there is no failover
mechanism. The application pick
the first target that it supports and uses that. There is no mention of trying
other records afterwards. Matching/finding NAPTR records stops once the first
match is completed. All other records are discarded. From section 2:
Hi Alex, two comments:
1) What you say does not explain my issue (as in any case TLS or TCP
have more priority in my NAPTR's than UDP). In fact, my exact problem
is that my Kamailio is not performing NAPTR query. 2) About your quoted text, indeed RFC 2915 says that.
But that just
means that, an application (in this case a SIP client) must take
*first* the valid NAPTR record with better "order" (lowest value). But
RFC 2915 is just about NAPTR, it can not mandate that applications or
protocols cannot try other NAPTR records as failover mechanism. Any
SIP client or proxy implementing NAPTR records does failover to the
next NAPTR record in case the first one failed (at least those I've
tested).
It seems that sip-router behaves exactly as you describe. This it, it
performs the NAPTR resolution, chooses the record with best 'order'
(for the supported transports) and just uses it (it would do SRV
failover and so, but would never try other NAPTR records with less
'order').
However I've seen other SIP stacks also performing failover by taking
next NAPTR records. But indeed, RFC 2915 does not mandate it, instead
it says MUST NOT. So you are right.
Regards.
--
Iñaki Baz Castillo
<ibc(a)aliax.net>