It should establish two dialogues.

  Tolga

> -----Original Message-----
> From: [EMAIL PROTECTED]
> [mailto:[EMAIL PROTECTED] Behalf Of
> [EMAIL PROTECTED]
> Sent: Wednesday, January 28, 2004 10:21 AM
> To: [EMAIL PROTECTED]
> Cc: [EMAIL PROTECTED]; [EMAIL PROTECTED]
> Subject: RE: [Sip-implementors] About RFC 3265 SUBSCRIBE/NOTIFY!
> 
> 
> What happens in the situation I describe below?
> 
> UAC         Proxy          UAS1           UAS2
> ----        -----          ----           ----
>  
> SUBSCRIBE
> ------------>   SUBSCRIBE
>                ------------->
>                 SUBSCRIBE
>                ---------------------------->
>                 200 To-tag:aaa
> 200 To-tag:aaa <------------
> <------------- 
> 
> NOTIFY To-tag:bbb
> <------------------------------------------
> 
> The NOTIFY arrives before the 200 from UAS2
> 
> Does that establish 2 dialogs? I think most implementations will 
> reject that NOTIFY with 481.
> 
> /Hisham
> 
> > -----Original Message-----
> > From: ext Paul Kyzivat [mailto:[EMAIL PROTECTED]
> > Sent: 28.January.2004 16:23
> > To: Khartabil Hisham (Nokia-TP/Helsinki)
> > Cc: [EMAIL PROTECTED]; [EMAIL PROTECTED]
> > Subject: Re: [Sip-implementors] About RFC 3265 SUBSCRIBE/NOTIFY!
> > 
> > 
> > I don't think the rate limiting helps with the problem that 
> > the original 
> > poster was concerned with. He seems to be concerned that the 
> > SUBSCRIBE 
> > might fork, and that NOTIFYs will come back from many 
> > different forks. 
> > If that happens, each will be independently rate limited, but the 
> > aggregate rate seen by the subscriber depends on the number of forks.
> > 
> > More below.
> > 
> >     Paul
> > 
> > [EMAIL PROTECTED] wrote:
> > > Every Event notification package needs to specify the rate of 
> > > notification. In most packages so far, that rate is limited 
> > to 1 NOTIFY 
> > > every 5 seconds.
> > >  
> > > Also, all these NOTIFY requests are within the same dialog 
> > established 
> > > by the SUBSCRIBE.
> > >  
> > > Regards,
> > > Hisham
> > > 
> > >     -----Original Message-----
> > >     *From:* [EMAIL PROTECTED]
> > >     [mailto:[EMAIL PROTECTED] 
> > Behalf Of *ext
> > >     Jerry Ipe Thomas
> > >     *Sent:* 28.January.2003 05:23
> > >     *To:* [EMAIL PROTECTED]
> > >     *Subject:* [Sip-implementors] About RFC 3265 SUBSCRIBE/NOTIFY!
> > > 
> > >     Hi Guys!
> > > 
> > >         In RFC 3265, it's said that if a proxy forks a SUBSCRIBE
> > >         request, then, one 2xx is returned to the subscriber for the
> > >         SUBSCRIBE request, but a helluva lot of NOTIFYs can 
> > be sent to
> > >         the subscriber!
> > >          
> > >         Does this mean this?
> > >          
> > > 
> > >             | USER_AGENT_SUBSCRIBER |                 | 
> > PROXY_SERVER |
> > >                     
> > SUBSCRIBE------------------------------->    ( Proxy
> > >             forks this to a lot of notifiers. This sounds 
> > like nonsense
> > >             to me anyway! )
> > >                         <----------------------------------2xx
> > >                         <---------------------------------NOTIFY
> > >                         <---------------------------------NOTIFY
> > >                         <---------------------------------NOTIFY
> > >                         <---------------------------------NOTIFY
> > >                                           .
> > >                                           .
> > >                                           .
> > > 
> > >                         <---------------------------------NOTIFY
> > >                        200---------------------------------->
> > >                        200---------------------------------->
> > >                        200---------------------------------->
> > >                        200---------------------------------->
> > >                                           .
> > >                                           .
> > >                                           .
> > >                        200---------------------------------->
> > > 
> > >         Aaaaaaaaaaaaaaah!!! God have mercy on the 
> > unfortunate subscriber!
> > >          
> > >         STUPID NOTE: The NOTIFYs are from notifiers.
> > >          
> > >         One possible way to get over this reeeeally scary 
> > scenario is
> > >         to match the subsequent NOTIFYs using the To tag in 
> > the 2xx to
> > >         the SUBSCRIBE??? Is this the right thing to do? Man! Forking
> > >         sucks! This IS a problem with devices that do not 
> > have much in
> > >         terms of processing horsepower or memory ( Say 33 
> > MIPS and 64KB
> > >         of SRAM. Remember that a lot of "other strange 
> > modules", which
> > >         are obviously not mine, are running within that 64KB. ) to
> > >         maintain separate dialogs with all those stupid notifiers!
> > >         I am pissed!
> > >          
> > >         It's obvious by now that I'm confused about forking and RFC
> > >         3265. :-(
> > 
> > The forking issues are similar in some respects to what they are with 
> > invite. If the proxy forks the SUBSCRIBE and then receives a 
> > 2xx it will 
> > return that to the subscriber. It will presumably then also 
> > attempt to 
> > CANCEL all the other forked legs. But there is a race 
> > condition, so some 
> > of the other forks may have already succeeded and sent 2xx responses 
> > back to the proxy. If so, the proxy is not permitted for 
> > forward them on 
> > the subscriber, so it just discards them.
> > 
> > Each leg of the fork that succeeded establishes a unique 
> > dialog, based 
> > on to & from tags. (To-tags are unique.) The subscriber learns about 
> > these dialogs from the messages it receives. Normally, the 2xx would 
> > establish the first dialog. For other legs, the 2xx isn't 
> > returned, so 
> > the first NOTIFY from the notifier on the leg tells the 
> > subscriber that 
> > there is another dialog.
> > 
> > If the subscriber only wants to hear from one notifier, it 
> > can terminate 
> > the extra dialogs by sending a reSUBSCRIBE w/Expires:0.
> > 
> > Yes, its a pain. Everything about forking is a pain.
> > 
> >     Paul
> > 
> > 
> 
> _______________________________________________
> Sip-implementors mailing list
> [EMAIL PROTECTED]
> http://lists.cs.columbia.edu/mailman/listinfo/sip-implementors
_______________________________________________
Sip-implementors mailing list
[EMAIL PROTECTED]
http://lists.cs.columbia.edu/mailman/listinfo/sip-implementors

Reply via email to