[SR-Users] uac module From header rewriting problems

[Timo Teras]

On Tue, 03 Jul 2012 16:23:45 +0200 Daniel-Constantin Mierla
<miconda at gmail.com> wrote:

> 
> On 7/3/12 3:49 PM, Timo Teras wrote:
> > On Tue, 03 Jul 2012 12:16:38 +0200 Daniel-Constantin Mierla
> > <miconda at gmail.com> wrote:
> >
> >> Hello,
> >>
> >> On 7/3/12 8:02 AM, Timo Teras wrote:
> >>> Hi,
> >>>
> >>> I fixed some uac From rewriting issues in commit e1d1c774c9ac0b4d,
> >>> however even with the latest versions I'm still seeing corruption
> >>> in From headers when they are rewritten for the whole dialogue
> >>> (from_restore_mode=auto, the default).
> >>>
> >>> I spent some time analyzing the problem and the problem seems to
> >>> be connected with the fact that only the bitwise difference of the
> >>> original and modified URI are stored (XOR).  The only purpose for
> >>> this appears to be the idea to save only the difference in the rr
> >>> params instead of two full URIs.
> >>>
> >>> However, it appears that in many cases (especially call transfer)
> >>> the From header will be altered during dialogue. This is
> >>> explicitly allowed in RFC 3261 § 20.20:
> >>>      The From header field indicates the initiator of the request.
> >>> This may be different from the initiator of the dialog.
> >>>
> >>> I've seen this in practise with few commercial SIP stacks.
> >>> Sometimes the changes are minor, or even trivial - but with the
> >>> difference encoding having even one character changed (e.g.
> >>> case), being deleted or added will result in the whole header
> >>> being corrupted.
> >>>
> >>> I'm now wondering if we should just store both URIs in the rr
> >>> params. While taking little more space, it should fix us sending
> >>> garbage to the wire (which usually results in dropped call as the
> >>> remote will reject the garbage messages).
> >>>
> >>> Any better ideas?
> >> it makes sense storing the entire value. Not sure if anyone else
> >> will want to be made configurable via mod param, I will go for
> >> simplest approach and have only the correct option, it will help
> >> in code maintenance.
> > Sounds good. Should probably do it for the stable branches too. I
> > can help with some of this, if needed. You have any estimate for
> > the fix?
> 
> I just expressed my opinion on how to fix it, but there is no plan
> for me to work on it in the near future, being caught in other tasks
> and traveling. So you can go ahead and make a patch to fix it for
> master, then later, based on the impact, it can be backported.

Ah. Ok. "I will go" just sounded like you'll do it. Sorry for
confusion. I'll try to take stab at it then when I get a minute for
it. :)

-Timo