Paul,

I really don't have a problem with a UAC behaving this way for an outbound
proxy. I'm more concerned with more complicated UAs like a gateway that
selects the proxy/redirect server based on things like the calling & called
numbers. I'm also wondering whether a proxy would have to behave the same
way if it inserted a Route header in order to direct the request to a
redirect server. Or in the case of a proxy would the "original request" be
the request is received (w/o the Route header to the redirect server)
instead of the request it sent to the redirect server?

cheers,
(-:bob

Robert F. Penfield
Chief Software Architect
Acme Packet, Inc.
130 New Boston Street
Woburn, MA 01801
[EMAIL PROTECTED]

----- Original Message -----
From: "Paul Kyzivat" <[EMAIL PROTECTED]>
To: "Bob Penfield" <[EMAIL PROTECTED]>
Cc: "Jonathan Rosenberg" <[EMAIL PROTECTED]>; "Magnus Edev�g"
<[EMAIL PROTECTED]>; <[EMAIL PROTECTED]>
Sent: Thursday, January 02, 2003 10:58 AM
Subject: Re: [Sip-implementors] request uri & where to send message?


> Bob,
>
> This depends on how the UAC is configured. Many can be configured with
> an "outbound proxy" thru which all outbound requests are routed. Based
> on 3261 it seems that the preferred way for the UAC to implement the
> outbound proxy functionality is to preload requests with a Route header
> referencing the outbound proxy. In general there will be little
> intelligence in the UAC about this - it will simply insert the Route
> header in all new outbound requests. (Presumably those that aren't part
> of a dialog.)
>
> It would be unwise to configure such a UAC with an outbound "proxy" that
> is in reality a server which always returns redirect responses. It
> clearly wouldn't work.
>
> You could of course construct a UAC that knows about a redirect server,
> and knows how to use it. But it would require more complex logic in the
> UAC to do the right thing.
>
> You can't simply expect a UAC that has an outbound proxy to bypass that
> proxy every time it gets a 3xx response. The 3xx may have come from
> further downstream, and the proxy may still be needed when retrying with
> the new address.
>
> Paul
>
> Bob Penfield wrote:
> > inline
> > ----- Original Message -----
> > From: "Jonathan Rosenberg" <[EMAIL PROTECTED]>
> >
> >>inline.
> >>
> >>Bob Penfield wrote:
> >>
> >>>inline
> >>>----- Original Message -----
> >>>From: "Paul Kyzivat" <[EMAIL PROTECTED]>
> >>>
> > <snip>
> >
> >>>>If the client is treating the server as an outbound proxy, by using a
> >>>>preloaded Route header, why would it decide not to use the same
> >>>>preloaded route header when sending to the redirected address?
> >>>>
> >>>
> >>>
> >>>You could certainly do that, but I if that server is just a redirect
> >>
> > server
> >
> >>>and not a proxy, there is no need to send the subsequent request back
> >>
> > thru
> >
> >>>the server.
> >>
> >>The UAC may not know that, though. Indeed, the spec does say that the
> >>Route headers would be present in the recursed request. From 8.1.3.4:
> >>
> >> > In all other respects, requests sent upon receipt of a redirect
> >> >   response SHOULD re-use the header fields and bodies of the original
> >> >   request.
> >>
> >>Local outbound redirect servers, IMHO, are a bad idea, because they can
> >>cause a ping-ponging effect.
> >>
> >
> >
> > This does not seem right to me. The UAC has decided by some local policy
to
> > contact a given next-hop server and sends it a request. That server
says, I
> > can't do it, but so-and-so can by sending a redirect response. Why would
the
> > UAC then send the request back to the same server that just told it to
go
> > somewhere else?
> >
> > With the above rule, how would you ever not send it back to the redirect
> > server? My understanding of 3261 is that unless you are sending the
request
> > to host indicated in the Request-URI, you put a Route header in.
> >
> >
>
>

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

Reply via email to