Olivier wrote:> Hi,
>
> I've seen some hardphones or Softswitchs now support this sip.instance
> feature :
>
http://www.softarmor.com/wgdb/docs/draft-jennings-sipping-instance-id-01.txt
>
> I don't really see any convincing use of this draft but I would be
> curious to share thoughts on it.
>
> Cheers
>
> ------------------------------------------------------------------------
We see similar things a lot from X-Lite (although their implementation
is somewhat different and has, in the past, been broken) -- using
rinstance or some such. The IDEA is sound to a point. It's useful to be
able to have an identifier with the UA to determine which UA is which
for purposes of differentiation if people are logging in multiple times
with the same username (something Asterisk doesn't allow).
It's usually handled with simple registration parameters using IP/port
combinations as differentiators, but if you're running a symmetric NAT,
that may be misleading or even non-functional. The instance COULD act as
an additional identifier to help clarify those situations on the SIP
side (as opposed to just the RTP side).
I do NOT, however, feel that such an identifier need last through
reboots or even be semi-permanent in any way. It's logically nice to
have a separation of instances of the same IP's clients (some of our
users log in multiple times from multiple machines from the same IP) for
programmatic purposes perhaps, but on a reboot, a new mapping should and
would be sent via the REGISTER request, and so keeping this data across
UA reboots seems.... unnecessary. And likely a security risk.
Where are you seeing this crop up?
N.