Transport-cc spec and implementation

[Boris Grozev]

Hello,

We are implementing the transport-cc extension and feedback mechanism,
and noticed a difference between the draft[0] and the webrtc.org
implementation.

For the 1-bit symbols in the RTCP feedback message the draft uses 0 for
"packet received" and 1 for "packet not received" (section 3.1.4). The
code, uses 0 for "packet not received" and 1 for "packet received"
(transport_feedback.cc, the EncodeOneBit function). This is a small
difference, but it breaks interoperability.

Since the draft is now expired (and also because what's in the code
seems to make more sense, matching the values of the 2-bit symbols), I
expect that the current code should be considered the reference in this
case. Is this correct? If so, are there any plans to update the draft?


Regards,
Boris

[0]
https://tools.ietf.org/html/draft-holmer-rmcat-transport-wide-cc-extensions-01

[Martin Storsjö]

Hi,

On Tue, 21 Mar 2017, Boris Grozev wrote:

> We are implementing the transport-cc extension and feedback mechanism, and
> noticed a difference between the draft[0] and the webrtc.org implementation.
>
> For the 1-bit symbols in the RTCP feedback message the draft uses 0 for
> "packet received" and 1 for "packet not received" (section 3.1.4). The code,
> uses 0 for "packet not received" and 1 for "packet received"
> (transport_feedback.cc, the EncodeOneBit function). This is a small
> difference, but it breaks interoperability.
>
> Since the draft is now expired (and also because what's in the code seems to
> make more sense, matching the values of the 2-bit symbols), I expect that the
> current code should be considered the reference in this case. Is this
> correct? If so, are there any plans to update the draft?

FWIW, even though that actual sentence seems to have it the wrong way
around, the example below seems to have it the right way, matching the
code.

// Martin

[Boris Grozev]

Ah, good point. So it appears it's just a typo in the draft.

Thanks,
Boris