inline On Wed, 2003-10-08 at 09:10, Ganesh Jayadevan wrote: > Chris, > > Thanks for the response. On reading the definition again, it does cover > it as you say. > > I would still like to suggest a change in wording for a couple of reasons: > > 1. The R-URI should be expected to be unchanged as loose routing gets > more and > more adopted. Therefore, this may not be typical. > 2. The example that follows is a carry over of the strict routing days. > > Spiralling should be expected to happen much more with loose routing > than before. > This clarification could be very useful. > > Is this the forum to request change or should I post it elsewhere? I've captured it.
Loop detection at proxies is also a concept that is a carryover from pre-3261 days. There are many reasons to not loop detect at proxies - the biggest being the n^2 processing burden it puts on the network. > > Thanks, > Ganesh > > Chris Boulton wrote: > > >Ganesh, > > > >I think the text already adequately covers this in the definitions section (6):- > > > >....but this time differs in a way that will result in a different processing > >decision than the original request. > > > >It does say that 'Typically' this will be a change in the R-URI BUT doesn't > >explicitly suggest this scenario. The fact that the route set has changed should > >be enough to differentiate from being a loop and fall into the category of being a > >spiral. > > > >Regards, > > > >Chris. > > > > > > -----Original Message----- > > From: Ganesh Jayadevan [mailto:[EMAIL PROTECTED] > > Sent: Tue 07/10/2003 18:09 > > To: [EMAIL PROTECTED] > > Cc: > > Subject: [Sip-implementors] when is a INVITE in a spiral (when not in a loop?) > > > > > > This isa repost. Sorry about the typo in the Subject line. > > > > Folks, > > > > I wonder if the definitions of Spiral and Loop needs to be updated. > > > > I am speaking within the context of using SIP in the 3G/UMTS architecture: > > > > S-CSCF or Serving call session control function is essentially a SIP proxy + > > more. > > > > A S-CSCF , can send an initial INVITE to a SIP application server (AS) for > > application > > processing. This can be done by loose-routing the request by including > > a set of one more route headers to the initial INVITE. > > > > If the AS is working as a proxy, it can manipulate the INVITE and send it > > right back to the S-CSCF. If the AS happened to be a loose router, it will > > not hammer the R-URI of the INVITE and instead will follow the route set as > > it sees it. > > > > The INVITE that comes back to the S-CSCF must not be thought of as being in > > a loop. It's route-set and Via headers, will be different and so the S-CSCF > > can > > distinguish it from the initially received INVITE. It will also fail the loop > > detection > > test as defined in section 16.6, step 8, 3rd paragraph (since the test > > includes > > hashing on the top-most Via). > > > > imho, an INVITE arriving at an proxy more than once has to fall into > > one of two buckets: Loop or Spiral. Currently I see the gap between the > > two caused due to R-URI being the same because of loose routing. > > > > What does the cognoscenti think of this? Should not the definitions of 'loop' > > and 'Spiral' be updated in Section 6? > > > > Thanks, > > Ganesh > > > > > > _______________________________________________ > > 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 _______________________________________________ Sip-implementors mailing list [EMAIL PROTECTED] http://lists.cs.columbia.edu/mailman/listinfo/sip-implementors
