Re: [MMUSIC] DECISION: Default mechanism to map RTP data to m- line is based on PT?

On 6/25/13 1:11 PM, Bernard Aboba wrote:
> Agree with Emil that "default" makes more sense for use of PT for mapping to an m line. If a=SSRC is specified on an m line then that should take precedence over PT.

I don't think "takes precedence" is quite the right way to think about it.

You support some number of mechanisms: PT only, or PT & SSRC, or ...

For the set of mechanisms you support, for each m-line you have some 
values for some of the mechanisms. When you receive a packet, attempt to 
match it to each m-line based on the the values it contains and those 
specified for each m-line. If a packet matches more than one m-line, 
then choose the m-line that matches the most values. If that still 
matches multiple m-lines, then there is an error.

You can statically analyze an offer and a candidate answer and determine 
if there can be any non-unique matches. If so, then you should not send 
that answer. Instead, send one that is more unique, or else reject some 
of the m-lines, or refuse to bundle some of them.

You can also statically analyze the O/A and determine a precedence for 
applying the mechanisms that minimizes the checking required for doing 
the matching. When using PT & SSRC, depending on the details it may be 
favorable to check one or the other first.

	Thanks,
	Paul

> On Jun 25, 2013, at 10:03 AM, "Emil Ivov" <emcho@jitsi.org> wrote:
>
>>
>>
>> On 25.06.13, 13:34, Christer Holmberg wrote:
>>> Hi,
>>>
>>> Emil suggested that the default, “MTI”, mechanism for mapping RTP data
>>> to m- lines should be based on PT.
>>
>> Actually, what I thought was important was having it as the "default" mechanism. Personally I don't mind having it as MTI might but it might be a tad too strong: implementations that deem PT demux insufficient and wouldn't want to use bundle without it, don't need implement it.
>>
>> We should probably also agree on how an answerer can determine whether or not it is expected to provide additional demuxing information (e.g. SSRCs) or if the offerer is using the default mechanism.
>>
>> One way would be for the answerer to inspect all m= lines and try to detect duplicating PTs. This could be a bit clumsy though. The offerer could expect SSRCs but its first offer may not yet contain PT reuse.
>>
>> With that in mind ... (more below)
>>
>>> Applications are allowed to use
>>> whatever other mechanisms, but usage of such mechanisms must be
>>> negotiated (or, applications need to have some other means knowing that
>>> the other endpoint support such mechanisms).
>>>
>>> *Q3*: Do we need to specify a default, MTI, mechanism for mapping RTP
>>> data to m- lines?
>>
>> I am in favour of having a "default" one.
>>
>> Without this BUNDLE would have to be an abstract spec and I don't see why we would want to leave it that way.
>>
>>> *Q4*: If Q3, do we mandate applications to support, and use (unless
>>> applications are made aware of other mechanisms supported by all
>>> endpoints) PT for mapping received RTP media?
>>
>> In favour.
>>
>> Cheers,
>> Emil
>>
>>>
>>> This means that, when mapping RTP to m- lines is required (whether it is
>>> always mandated is discussed in another thread), within a BUNDLE group
>>> each PT value must be unique to an individual m- line.
>>>
>>> NOTE: If your answer to Q3 is “yes”, but your answer to Q4 is “no”,
>>> please indicate which mechanism you prefer J
>>>
>>> Regards,
>>>
>>> Christer
>>>
>>>
>>>
>>> _______________________________________________
>>> mmusic mailing list
>>> mmusic@ietf.org
>>> https://www.ietf.org/mailman/listinfo/mmusic
>>
>> --
>> https://jitsi.org
>>
>> _______________________________________________
>> mmusic mailing list
>> mmusic@ietf.org
>> https://www.ietf.org/mailman/listinfo/mmusic
> _______________________________________________
> mmusic mailing list
> mmusic@ietf.org
> https://www.ietf.org/mailman/listinfo/mmusic
>