Rohan,

that is surely going into the right direction!

I think it clearly points out the need to transport more than one
address in the SDP.

However, it does not seem to solve the problem for requests that don�t
have SDP attachments (like MESSAGE). There is a difference in SIP
routing and RTP routing. I think the draft addresses the RTP routing
problem. For SIP routing, I think we still need a way for formulating
that non SIP-aware equipment is involved in the routing process.

Again, in my perception, SIP deals with *routes* (user agent client,
number of routers with loose/strict/no routing and the user agent
server). Every router has a defined behaviour (SIP aware called
"proxies" and non SIP-aware called "NAT"). We know that the different
types of proxies need different treatment and I begin to realize that
NAT routers need a different treatment as well.

The good news is that "no-routers" can be treated backward compatible.
If the routing element behind the no-router fixes the SIP routing
information everything should run fine. It is also the element behind
the no-router that inserts this information. 

Anyway, having a STUN server on board of every SIP network element
(UA/proxy) is probably a good idea for whatever comes. As you know, we
did this already half a year on our proxy!

Best,


Christian

> -----Urspr�ngliche Nachricht-----
> Von: Rohan Mahy [mailto:[EMAIL PROTECTED]
> Gesendet: Samstag, 22. M�rz 2003 05:57
> An: Christian Stredicke
> Cc: [EMAIL PROTECTED]; 'J�rgen Bj�rkner'; 'Robert
Sparks'
> Betreff: Re: "Strict Routing", "Loose Routing", "No Routing"!
> 
> Hi Christian,
> 
> Have you seen Jonathan Rosenberg's "ICE" proposal which we discussed
in
> SIPPING?  How do you think these would relate.
> 
> thanks,
> -rohan
> 
> 
> On Monday, March 17, 2003, at 09:17 PM, Christian Stredicke wrote:
> 
> > Hello again,
> >
> > while we have now loose routers and strict routers it seems that a
new
> > type occurs on the horizon: The "no router"!
> >
> > This is for example the case when using NAT forwarding (e.g. with
STUN
> > or UPnP). The router is not at all aware of SIP and does not
manipulate
> > the packet at all. However, the router is too dumb to forward a
packet
> > if it is coming from the wrong direction. This is the case with most
of
> > the DSL routers today.
> >
> > May I propose to use a ";nr" flag in a URI to indicate that the
router
> > does not do any SIP changing at all?
> >
> > User Agents that are able to deal with "no routers" would then
register
> > with the following packet:
> >
> > REGISTER sip:bla.com SIP/2.0
> > Path: <sip:62.12.245.32:54456;nr=1>
> > Contact: <sip:[EMAIL PROTECTED]:5060>
> > ...
> >
> > When it receives a message this could look like this:
> >
> > INVITE sip:62.12.245.32:54456;nr=1 SIP/2.0
> > Route: <sip:[EMAIL PROTECTED]:5060>
> > ...
> >
> > SIP/2.0 200 Wonderful
> > Record-Route: <sip:62.12.245.32:54456;nr=1>
> > Contact: <sip:[EMAIL PROTECTED]:5060>
> > ...
> >
> > The nr parameter could easily indicate that the previous router did
not
> > change the packet at all and that the next entity has to fix this
> > before
> > processing the message. Alternatively, the router sending the packet
> > could fix the route before sending it.
> >
> > What do you think? Is this really necessary? Just omit the nr
parameter
> > and conclude implicitly? How else should we deal with routers that
are
> > not SIP-aware? I don�t like the route encoding in the header
> > (sip:abc?route=def) as it raises many questions about how to merge
this
> > with other existing routes.
> >
> > Unfortunately, omitting the private address is no solution, as
devices
> > in the same subnet have to bypass the "no router". Many DSL routers
are
> > not able to send packets from private address via the public address
to
> > the other private address.
> >
> > Cheers,
> >
> >
> > Christian
> > --
> > Dr. Christian Stredicke
> > sip:[EMAIL PROTECTED]
> >



_______________________________________________
Sip-implementors mailing list
[EMAIL PROTECTED]
http://lists.cs.columbia.edu/mailman/listinfo/sip-implementors

Reply via email to