Hi,

Yes UA1 does send an ACK to ser, but the request URI contains the address of UA3 and the route heard contains the address of ser. With this scenario, ser routes the request to the initial called UA (UA2) instead of going to UA3. The second ACK message still coming from UA1 (but a different UA, ATA186 in this case) does generate an ACK but in the request URI it's the ser proxy address, and in the route header it's the address of UA3, and in this case ser route the ACK to UA3.

In both cases the ACK is generated from UA1 (but they are different brands) and they go to ser.

Included are two ethereal files, one is redirect-fail which contain the redirection to a host @ port 5071 (UA3) while the second one redirect-ok still contain the same redirect but this time it's working.

Phoenix.tunix.com:5071 is the address of UA3 (asterisk).

Thanks.

Samy.

-----Original Message----- From: Jiri Kuthan [mailto:jiri@iptel.org] Sent: Sunday, February 15, 2004 3:42 PM To: Samy Touati; serusers@lists.iptel.org Subject: Re: [Serusers] UAs not respecting Record Route

At 08:46 PM 2/15/2004, Samy Touati wrote:

Hi,

I do have the following situation:

UA1 calls UA2 going through ser 0.8.12. All signaling is Recourd Routed and

go through ser.

UA2 doesnt answer and I do have a t_on_failure to go to UA3. UA3 receives INVITES from ser, and answer with a 200/OK with sdp, this

messages goes through ser which forward it to UA1.

UA1 sends an ACK to ser having this in message:

Session Initiation Protocol Request line: ACK sip:ua3@22.2.4.193:5071 SIP/2.0 Method: ACK Message Header Via: SIP/2.0/UDP 2.20.64.193:5063;branch=z9hG4bk-7f415771 From: test1 sip:ua1@vocal.ipsound.net;tag=eb646133bf338b12 To: sip:ua2@vocal.ipsound.net;tag=as47221fa0 Call-ID: a66804e6-5ac924fd@142.133.80.102 CSeq: 102 ACK Max-Forwards: 70 Route: sip:sip-proxy@ser-domain.com;ftag=eb646133bf338b12;lr=on Proxy-Authorization: Digest username=" Contact: ua1 sip:ua1@24.202.64.193:5063 User-Agent: Sipura/SPA2000-1.0.10 Content-Length: 0

Ser is supposed to follow the route, and ends up looking into the location

table and re-forwards the request to UA2 again (even if it already timed out once).

Interesting -- it looks like properly generated ACK, except the URI in Route has some strange value -- I mean SER generates the URI as sip:<username_from_INVITE>@<SER_IP>;ftag=... How did these values appear there?

Are you sure your script does include the initial if (loose_route) { t_relay();break; } sequence? That should forward the request to address in request URI.

Now I do have another UA which in this same scenario sends the correct ACK:

Session Initiation Protocol Request line: ACK sip:sip-proxy@ser-domain.com SIP/2.0 Method: ACK Message Header Route: sip:ua3@22.2.4.193:5071 Via: SIP/2.0/UDP 2.20.64.193:5063 From: ua1 sip:ua1@ser-domain.com;user=phone;tag=3551608275 To: sip:ua2@ser-domain.com;user=phone;tag=as112d04ea Call-ID: 2653016335@142.133.80.101 CSeq: 2 ACK User-Agent: Cisco ATA 186 v3.0.0 atasip (031210A) Proxy-Authorization: Digest username Content-Length: 0

In this case ser follow the route header, and ends up sending the ACK to

UA3.

Snom phone with 2.03o firmware as well as sipura with 1.0.10 and 1.0.29b

firmware have the same wrong behaviour.

What is the right way of having this ?

The problem with snom may be the phone uses some preloaded Routes which are incompatible with some record-routing workaround for a Cisco gateway problem in SER. I remember it was reported on the archive -- you perhaps find more there.

My understanding is that in record route situation, ser looks into the

route header, and sends the message based on it.

yes

Here I do have User Agents which seem to create an ACK message ignoring it

should go to a sip proxy .

You told above UA1 does send to proxy, did not you? It would be better to send a complete message dump and config file.

-jiri