Jeroen van Bemmel wrote:
Paul,

I don't see any reason why *proxy* middle boxes need to know anything about this. B2BUAs will need to know, but then they will know.

There are (as always) subtle differences. NOTIFY for SUBSCRIBE is needed to handle forking (proxies pass at most 1 2xx response for SUBSCRIBE), while NOTIFY for INVITE would not be needed for this reason. Proxies Record-Route NOTIFY due to the above reason, they could avoid doing that for INVITE-NOTIFYs.

I don't follow. If the proxy was going to R-R, it needed to do so in the SUBSCRIBE. (Or it wouldn't even see the NOTIFY.) If its doing asymmetric R-R then it may need to update the R-R entry, but that is conditional on the entry it put there in the SUBSCRIBE. The UAS for the NOTIFY needs to recognize if it establishes a new dialog and if so remember the R-R, but again that doesn't impact a proxy.

Record-Routing INFO is useless and somewhat wasteful, so well designed proxies don't do that.

Also, I assume there will be no need to "immediately send NOTIFY" for INVITE-based events.

True. But that again is a matter for the UAC and UAS, not the proxies.

All in all, IMO using INFO instead of NOTIFY here is preferrable (also given that existing implementations already use INFO within INVITE dialogs).

No matter what we do, the existing standardized uses of INFO will remain.

And I think it is safe to assume that the existing *unstandardized* uses of INFO will not do the kind of negotiation we are talking about, so they won't carry over directly.

So we are really only talking about new things.

If we use NOTIFY, then existing, already standardized event packages can (hopefully) be easily adapted to support the new technique as well as the old. Because the semantics of the package wouldn't change, I would hope that it would be easy to get those changes adopted.

If we use INFO, then I guess we can hopefully easily adapt existing non-standard INFO usages. But because they haven't been standardized, that may be more difficult.

I suppose one could adapt existing event packages to use INFO for this style and NOTIFY for freestanding subscriptions. Mechanically its not a big deal, but it certainly seems odd.

        Thanks,
        Paul


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

Reply via email to