Re: [rtcweb] Call for adoption of QoS draft

[James Polk <[e-mail]>]

At 04:27 PM 9/10/2012, Martin Thomson wrote:
>On 10 September 2012 14:12, James Polk <[e-mail]> wrote:
> > MMUSIC has a WG item that should provide this indication/hint, in a
> > trafficclass label attribute. Browsers identifying a small set of labels
> > from that effort should do the trick. See
> >
> > 
> http://www.ietf.org/internet-drafts/draft-ietf-mmusic-traffic-class-for-sdp-02.txt
>
>Thanks James, that looks like what is needed.  It's not exactly the
>same thing as the priority draft, so I'm not sure how I'd map all the
>way from priority (1-3) to traffic class labels to the DSCP markings
>described in Cullen's draft.

I'm not sure only 3 priorities are needed, though I believe 3 will 
get used. The point of the trafficclass labels (TCL) is to add a 
level of abstraction to what the actual DSCPs need to be by allowing 
each domain that is aware of the type of traffic choose the 
appropriate DSCP for that type of traffic. After all, some bright 
spark might just come along and suggest that even though RTCweb uses 
the same 5-tuple for all traffic, they propose that the audio packets 
use a different DSCP than the video packets.

Further, someone might consider the type of AV conference they are in 
to determine which DSCP they are to use. Say the difference between a 
small window av call and an immersive telepresence room experience? 
Both types of calls have the exact same types of packets (audio and 
video, and maybe preservation sharing or whiteboarding).

Another is Dan's point in which Alice uses her highest priority 
marking (DSCP 1) when she contacts Bob, who uses his highest priority 
marking (DSCP 2) for the same call. Now, which DSCP is the highest 
priority marking, 1 or 2? Does it depend on the network/domain the 
packet is in?

The TCL draft explores that, and we already have customer feedback 
that's caused us to allow greater flexibility than we originally planned.


>It looks like some more definition work is needed.

perhaps, and I'm not at all opposed to comments to the TCL draft 
within MMUSIC to address whatever folks think is important.

James