Re: [Sip] Ensuring in-dialog stuff works with outbound

[Rohan Mahy <rohan at ekabal.com>]

Hi Byron,

Double record-route is not needed.  A simple Record-Route with a flow  
token in the userpart is sufficient from an implementation  
perspective, as long as the edge proxy can use the source of a  
request to determine if a request routed through a particular flow  
token is "coming" or "going".

Also, the draft just says that each edge proxy just needs to do  
whatever is necessary to make sure that subsequent requests get back  
to the same instance.  This leaves the door open for 1) extra edge  
proxies between the first edge proxy and the authoritative proxy to  
step out of the path, 2) edge proxies that record route with a more,  
3) incorporating per call state blobs, (and possibly other things I  
haven't thought of).

thanks,
-rohan

On Apr 20, 2007, at 9:33 AM, Kevin Johns wrote:
> Byron,
>
> This was along the lines of what I was thinking, but neither outbound
> nor RFC 3327 state this.
>
> Kevin
>
> -----Original Message-----
> From: Byron Campen [mailto:bcampen at estacado.net]
> Sent: Friday, April 20, 2007 9:08 AM
> To: Kevin Johns
> Cc: SIP IETF; Cullen Jennings; Rohan Mahy
> Subject: Re: [Sip] Ensuring in-dialog stuff works with outbound
>
> 	Yeah, this is essentially what I was asking. I ended up having a
> conversation with Rohan, and the takeaway was that the outbound-  
> enabled
> UA needed to remember the edge-proxy's Path header, and pre- load  
> it as
> a Route header in outgoing requests over that flow. The edge-proxy,  
> when
> seeing this, would then Record-Route with the same (after determining
> whether this flow-token referred to the source, or the destination, it
> would Record-Route in such a way to keep it from getting confused over
> the direction the flow-token was pointing, probably by using a double
> record-route.)
>
> Best regards,
> Byron Campen
>
> > Byron,
> >
> > Just saw this email and have not seen any responses as of yet...
> >
> > I am not sure I fully follow your issue, are you trying to figure out
> > how to distinguish between non-outbound enabled UAs and outbound
> > enabled
> > UAs within the same edge proxy? It also appears you want to be  
> > able to
> > determine this so you know when to insert a flow token in a record-
> > route
> > header to allow for routing of mid-dialog requests?
> >
> > Is my understand correct?
> >
> > Kevin
> >
> >
> > -----Original Message-----
> > From: Byron Campen [mailto:bcampen at estacado.net]
> > Sent: Thursday, April 05, 2007 1:00 PM
> > To: SIP IETF
> > Cc: Cullen Jennings; Rohan Mahy
> > Subject: [Sip] Ensuring in-dialog stuff works with outbound
> >
> > 	I've been tuning a server implementation of outbound recently,
> > and I've come upon a bit of a conundrum. Say we have an endpoint  
> > using
> > outbound with an edge-proxy. Everything is set up as specified in the
> > draft. This endpoint then sends a dialog-forming request. The edge
> > proxy
> > has to be able to ensure that in-dialog traffic coming from the other
> > direction goes over the same outbound flow, by record-routing with a
> > flow token. However, in the non-outbound case, is it downright  
> > _wrong_
> > for the edge proxy to record-route with a flow-token, because  
> > doing so
> > breaks target-refreshes. Unfortunately, the edge-proxy is keeping no
> > flow state, and has no clue if this dialog-forming request came
> > over an
> > outbound flow or not.
> >
> > I see two choices for fixing this, please pipe up if you see
> > additional
> > options:
> >
> > 1. We require the endpoint to specify whether it wants the outbound
> > treatment on each outgoing request, either by including some
> > recognizable outbound goo in the request, or by preloading a Route
> > header with a flow-token to the edge proxy (using something like
> > Service-Route maybe, although interactions between Service-Route and
> > Path might be a little weird).
> >
> > 2. We require proxies to store information on which connections are
> > outbound flows, and which are not (this would make the whole
> > exercise of
> > stateless flow-tokens moot).
> >
> > Best regards,
> > Byron Campen



_______________________________________________
Sip mailing list  https://www1.ietf.org/mailman/listinfo/sip
This list is for NEW development of the core SIP Protocol
Use sip-implementors at cs.columbia.edu for questions on current sip
Use sipping at ietf.org for new developments on the application of sip