Re: [rtcweb] Offer/answer for heterogeneous encode/decode

[Harald Alvestrand <[e-mail]>]

On 11/23/2013 01:20 AM, Paul Giralt (pgiralt) wrote:
> On Nov 22, 2013, at 7:06 PM, Martin Thomson <[e-mail]> wrote:
>
>> On 22 November 2013 15:44, Paul Giralt (pgiralt) <[e-mail]> wrote:
>>> While you’re right that it would work for this scenario, my point was really
>>> that Offer/Answer is not really asymmetric as implied earlier. Take for
>>> example the hypothetical case where you are only able to decode VP8 but only
>>> able to encode H.264. If I offer VP8 as my only codec (because it’s the only
>>> thing I’m able to decode therefore I never want anyone to send me anything
>>> other than VP8) I cannot send H.264 in the offer because that implies I’m
>>> able to decode it. The other side then wants to say it can only receive
>>> H.264 so it would have to send back an answer with only H.264. I guess
>>> there’s nothing really inherently stopping you from doing this because as
>>> far as I can tell, 3264 only says the answer has to be a subset of the offer
>>> for multicast streams, however how would the answering side know that the
>>> offering side is even capable to receiving H.264? Perhaps Offer/Answer can
>>> technically be asymmetric, but it doesn’t seem practical to use it this way
>>> because you cannot really indicate your send and receive capabilities
>>> independent of each other.
>> Judicious application of a=sendonly or a=recvonly avoids this issue.
>> If you want to send H.264, try a=sendonly on a line with H.264.  If
>> you want to receive VP8, try a=recvonly on a line with VP8.
> I gave this as an example of a possible way to do it, but that means you need two separate m= lines - one for send and one for receive. Seems strange to do this for something that you really want to be a single bi-directional stream.

An application that can't talk to itself is kind of bizarre too.

I think this is a corner case.