Re: "Priority and QoS model"


> On Jun 18, 2017, at 12:39 AM, Stefan Håkansson LK <stefan.lk.hakansson@ericsson.com> wrote:
> 
> Hi all,
> 
> the section "Priority and QoS model" [1] basically gives the JS 
> application the options "very-low", "low", "medium" and "high" and then 
> references RTCWEB-TRANSPORT section 4 and TSVWG-RTCWEB-QOS.
> 
> RTCWEB-TRANSPORT section 4 in turn describes "local prioritization" and 
> use of DSCP (with references to TSVWG-RTCWEB-QOS for the later).
> 
> Both "local prioritization" and DSCP use are phrased as "SHOULD". This 
> means that there may be no local prioritization and/or no DSCP marking 
> made even though a specific priority is requested, and the application 
> would not know.
> 
> I think this is what we agreed back when this was discussed, but I just 
> wanted to check that the group thinks this is fine.
> 
> Stefan
> 
> 
> [1] http://w3c.github.io/webrtc-pc/#priority-and-qos-model

> 

Yes, though not ideal, I think that is where we are. Part of the issue is that it might be hard to implement in some operating systems. The other part of the issue is even if the browser / OS did exactly the right thing, the network might remove the DSCP marking and the application will never know. Application that use this have to be prepared for the fact it will not work. 

Testing as part of the Cisco/Apple Fastlane project, which is basically just these DSCP markings, shows this makes a significant difference in many use cases.

Received on Tuesday, 20 June 2017 18:30:46 UTC