RE: [Sip] draft-ietf-sip-refer-02 - SHOULD strength when sending NOTIFY

[Robert Sparks <[e-mail]>]

inline

> -----Original Message-----
> From: JF Rey [mailto:[e-mail]]
> 
> Hi,
> 
> I'm a bit confused about the way NOTIFY are used by the 
> "referee" party.
> 
> referer              referee
>  |====== SIP session ===|
>  |--------REFER-------> |
>  |<--------202--------- |----INVITE -->
>  |                      |<---180------
>         ... quite a long time ...
>  |<------NOTIFY ------  |<---200------ 
>  |----------200-------->|
>  |---------- BYE ------>|
> 
> draft-ietf-sip-refer-02 states :
> "3.5.3.1 Using NOTIFY
> Once it is known whether the reference succeeded or failed, 
> the UA receiving the REFER SHOULD notify the agent sending 
> the refer using the NOTIFY mechanism [...]"
> 
> All the example flows I've seen around use the NOTIFY 
> mechanism whenever a REFER is accepted. 
> 
> If I understand the "SHOULD" strength correctly, the referer 
> can't know whether the referee will send a NOTIFY. So it 
> can't know when to stop waiting for a NOTIFY and when to send 
> a BYE. I think that's quite awkward. 

The referer is going to have to be prepared to never receive
a NOTIFY whether that requirement is a MUST or a SHOULD. IIRC, the
proponents of SHOULD earlier argued that they had systems where
they knew the referer did not care about the result and didn't want
the referee/network to be burdened with an effectively useless 
NOTIFY. However - later in the message you caused me to realize
we need to realign with sip-events. See below.

> 
> Besides, I understand that nothing prevents a referee to send 
> NOTIFY on provisional responses (and not only on final 
> responses). But the way the sentence is written : "whether 
> the reference succeeded or failed" does not really take 
> provisional responses into account. No example in 
> draft-ietf-sip-cc-transfer-05.txt shows NOTIFY triggered by 
> 1xx. If the referee INVITEs a human, the referer may have a 
> very long time to wait before receiving a NOTIFY triggered by 
> the 200 response to the INVITE.

This clarification has already been requested and will be added
to the drafts. I don't understand how the last sentence relates
to the rest of the paragraph though. It is correct in the presence
or absence of NOTIFYs carrying progress information.

> 
> Do people implement REFER without the NOTIFY mechanism ?

I don't know. If someone does it would help if they speak up.

>
> If 
> we can't change the SHOULD strength, could we have a way for 
> the referee to indicate that it will send NOTIFY ? 

The current sip-events draft requires an immediate notify,
so I think this may be moot. I think we need to align to that
change in sip-events - and the SHOULD you object to effectively
goes away. However, when implementing, you still need to protect
yourself against a NOTIFY never arriving for robustness.

>Could the 
> wording in 3.5.3.1 include NOTIFY on provisional responses ?
Yes


RjS

_______________________________________________
Sip mailing list  http://www1.ietf.org/mailman/listinfo/sip
This list is for NEW development of the core SIP Protocol
Use [e-mail] for questions on current sip
Use [e-mail] for new developments on the application of sip