[Serusers] Session-Expires Header

[Jiri Kuthan]

At 12:14 AM 3/9/2005, Zeus Ng wrote:
>Yes, you do need UA supporting this header.
>
>According to section 8 of the draft, draft-ietf-sip-session-timer-14:
>   Session timers are mostly of interest to call stateful proxy servers
>   (that is, servers that maintain the state of calls and dialogs
>   established through them). However, a stateful proxy server (that is,
>   a server which is aware of transaction state, but does not retain
>   call or dialog state) MAY also follow the rules described here.
>   Stateless proxies MUST NOT attempt to request session timers. Proxies
>   that ask for session timers SHOULD record-route, since they won't
>   receive refreshes if they don't.
>
>Here, SER is not a call stateful proxy but transaction stateful proxy. It
>will not maintain the state of the call. So, even if you insert the header,
>it will not change anything from SER's perspective.
>
>However, if as least one of the UA support this header, you may see a BYE
>request, which is good for accounting. If neither UA support the header, no
>re-INVITE or UPDATE will be generated and that doesn't help you more than
>what you have at the moment.
>
>So, until SER becomes call stateful (which is similar to B2BUA, but not the
>same), your best hope will be UA that support this header.

since the case you are raising is accounting accuracy, a pstn gateway is typically
invoivled and it then fair to assume that >=1 UA supports ST. If not, junk the
gateway :)

-jiri