-----Original Message-----
From: [email protected]
[mailto:[email protected]] On Behalf Of Dale Worley
Sent: Friday, August 07, 2009 12:07 PM
To: Keith Gearty
Cc: Bogdan Brezoi; [email protected]
Subject: Re: [sipx-users] CODEC scenario

On Fri, 2009-08-07 at 17:49 +0100, Keith Gearty wrote: 
> But that would be true if you followed my suggestion as well.  No
> codec manipulation is required, just filtering of the codec names
> (treated as text strings) in the SIP messaging, according to an
> exclude list.  That to me seems like the cleanest way to prevent
> high-bandwidth codec being used across a particular SIP trunk or
> site-to-site (gateway) interface.  If you can suggest a better way of
> achieving this, then please do.

Scott includes "editing SDP" in the category "codec manipulation".
Clearly, the cleanest way to restricting the codecs used with an ITSP is
to have the ITSP offer/accept only the codecs that the customer wants.
Since the ITSP is almost certainly encoding/decoding RTP, it is already
doing SDP negotiation.

That doesn't work if the two ends of the link in question are sipXecs's.

If one must edit SDP, problems will likely ensue.  It does seem that the
structure of SDP allows us to find all mentions of codecs that can be
used in the RTP, even if the endpoints are using SDP extensions that we
do not know.  A particular problem is that in order to enforce a
bandwidth restriction, the SDP editing must remove any codecs that the
filter does not understand.  This automatically interferes with
endpoints introducing new codecs.  This approach generates endless
ramifications that must be dealt with.

Here is an alternative approach that would be much more robust, although
I don't know if anyone is implementing it:  Have the border router
throttle the bandwidth consumed by each individual IP endpoint.  (I'm
sure that there are routers that can do this.)  Then have endpoints
negotiate a set of codecs with differing bandwidths, and automatically
switch to a codec with a lower bandwidth if it detects RTP packet loss.
As far as I can see, this is within all existing protocol specs and
since it's negotiated end-to-end, it takes into account all sources of
network constriction.

Dale



Something in Dales discussion here got my attention.  IF the endpoints are a
telephone and a sipXecs system, does that mean there won't be a negotiation
of the speed then?  Can someone explain how that works technically.  One
example that I'm curious about is if you have two sipXecs located at two
different locations.  When a call is placed from the far end system to the
near end system auto attendant as an example, what Codec is use - I'm
assuming G711.

It could be a great feature for sipXecs systems to have a special
configuration for when they are communicating to another sipXecs system,
allowing them to have some features not available to other systems, and
giving some enhancements to the network itself - for example - additional
codecs like g729, etc.  Knowing the two endpoints systems are the same could
allow for some transferring of alarms, errors, etc. between the devices with
some specialty tunnels, etc. for a common user interface, centralized
management, etc.

Having not worked with the HA versions, this might be in there already and
I'm not aware.

Regards,

Todd Hodgen


_______________________________________________
sipx-users mailing list [email protected]
List Archive: http://list.sipfoundry.org/archive/sipx-users
Unsubscribe: http://list.sipfoundry.org/mailman/listinfo/sipx-users
sipXecs IP PBX -- http://www.sipfoundry.org/

_______________________________________________
sipx-users mailing list [email protected]
List Archive: http://list.sipfoundry.org/archive/sipx-users
Unsubscribe: http://list.sipfoundry.org/mailman/listinfo/sipx-users
sipXecs IP PBX -- http://www.sipfoundry.org/

Reply via email to