What vexes VoIP users?

Michael Thomas mike at mtcc.com
Tue Mar 1 21:53:26 CST 2011


On 03/01/2011 07:33 PM, Jay Ashworth wrote:
> Is IP running over an L2 with a SLA any less "IP" than one
>> without a SLA? That's all the DOCSIS qos is: dynamically
>> creating/tearing down enhanced L2 qos channels for rtp
>> to run over. It's been quite a while since I've been involved,
>> but what we were working on with CableLabs certainly was
>> VoIP in every respect I can think of.
>>      
> Wow.
>
> I thought I was pretty clear in what I said above; I'm sorry you didn't
> get it.
>
> "What everyone is actually *selling* commercially, except for cable
> providers, is *not* VoIP; it's a subset of that: VoN; Voice Over Internet;
> where the IP transport *goes over the public internet*, and through
> whatever exchange points may be necessary to get from you to the
> provider.
>
> Cable companies are selling you *one hop* (maybe 2 or 3; certainly not
> 12-18), over a link with bandwidth protected from whatever may be
> going on on the Internet IP link they're also selling you; and which is
> therefore guaranteed to have better quality than whatever "VoIP" service
> it might be competing with."
>
> Better?
>    

Uh, I was part of the standards at packetcable from around
1999 or so on, and it was always plain old rtp over docsis. Since
the upstream was really lousy back then, the MSO's wanted to
give committed bit rate l2 docsis channels to the rtp traffic,
but it was still rtp/rtcp flowing through them.

So I still have no idea what distinction you're trying to draw.

> I wasn't suggesting QOS.  I was suggesting *there's a completely separate
> pipe*, on non-Internet connected IP transport, carrying only the
> voice traffic, directly to a termination point, which is dedicated
> from the triple-play box and nailed up.
>
> Are you suggesting that's *not* how it's being done in production?
>    

There were some MSO's who were thinking about doing that,
but as I recall they went the way of the AAL2 dodo bird. Maybe
a few deployed it, but from a packetcable/cablelabs perspective
they weren't on the table. MGCP was the answer to getting
rid of class 5 switches altogether, which the MSO's didn't
have any particular affinity to. It was always rtp over ip over
DOCSIS with DSCP in the core and arguments about RSVP.

Mike, member of the packetcable security spec team whose
   work spawned SRTP and KINK amongst other things




More information about the NANOG mailing list