Re: [rtcweb] Replacing a=extmap mapping in re-offer, is it legal?

[Christer Holmberg <[e-mail]>]

Hi,

Taylor's explanation makes sense. But, in this case:

1) Re-using the same value with a new mapping should be an explicit MUST NOT

2) The re-offer should be rejected

Regards,

Christer

-----Original Message-----
From: Iñaki Baz Castillo [mailto:[e-mail]] 
Sent: 07 August 2017 19:56
To: Taylor Brandstetter <[e-mail]>
Cc: Christer Holmberg <[e-mail]>; [e-mail]
Subject: Re: [rtcweb] Replacing a=extmap mapping in re-offer, is it legal?

On Mon, Aug 7, 2017 at 7:46 PM, Taylor Brandstetter <[e-mail]> wrote:
> I don't know if this is stated explicitly anywhere, but my assumption 
> was that it works like payload types. You can introduce new IDs all 
> you want, but you can't have one ID refer to multiple things during the same session.
> Otherwise, if you receive a packet with that extension ID, you 
> wouldn't be able to tell which extension it is.
>
> Meaning, if Bob's initial answer was:
>
> a=extmap:11 http://www.webrtc.org/experiments/rtp-hdrext/abs-send-time
>
> It would be illegal to have a re-offer of:
>
> a=extmap:1 http://www.webrtc.org/experiments/rtp-hdrext/abs-send-time
> a=extmap:11 urn:ietf:params:rtp-hdrext:toffset
>
> Because if Bob then receives a packet with extension ID 11, Bob 
> doesn't know if it's a packet sent before Alice received the answer 
> (in which case it's "abs-send-time"), or after (in which case it's "toffset").

Thanks, your rationale clearly makes sense. However:

1) I don't think Firefox cares about overriding a previous extmap ID during a reoffer.

2) Should we really accept so complex and error prune things just because SDP allows everything?


--
Iñaki Baz Castillo
<[e-mail]>