Re: [rtcweb] Comments on draft-jennings-rtcweb-qos (Re: Call for adoption of QoS draft)

[Martin Thomson <[e-mail]>]

On 17 September 2012 15:16, Cullen Jennings (fluffy) <[e-mail]> wrote:
>> What do people think about NOT including QCI/WiFi markings?  Is it not
>> possible for a wireless interface to examine DSCP markings in order to
>> determine the markings for the link?  We should endeavour to maintain
>> the proper abstractions.
>
> On some OS is it is possible to see them but ignoring that … think for a second  about what these marketing are meant to be used for -  they are more meant to be seen by the network not the browser at the far end so not sure this matters much if the some OS can't provide this information to the application

I am not suggesting anything about reading the markings off packets.
Reading markings off is - as you say - close to useless.

I asked if it were possible to consider *not* specifying what link
layer markings are made for packets that are *sent*.  The suggestion
was that if DSCP markings were made, as suggested, the packets could
be marked on the way onto the link by the wireless driver using
whatever link-specific markings are appropriate.  That might lead to a
mapping from DSCP to QCI and WiFi markings, but not necessarily.  That
would be different to the straight-from-the-top application priority
to link-layer, as described in the draft.