Re: [MMUSIC] ice-options attribute: session- and media-level?

[Peter Saint-Andre <[e-mail]>]

On 6/19/18 5:37 PM, Paul Kyzivat wrote:
> Peter,
> 
> On 6/19/18 3:40 PM, Peter Saint-Andre wrote:
>> On 6/19/18 12:38 PM, Christer Holmberg wrote:
>>> Hi,
>>>
>>> The draft defines the ice-options SDP attribute, so the draft needs
>>> to define what type (media and/or session) of SDP attribute it is.
>>
>> At the syntax level, yes. Sematically, however, is it perhaps the job of
>> the ICE specification to define whether ICE extensions (however encoded
>> in a using protocol) are session-level or media-level?
> 
> whether session level, media level, or both is required as part of the
> registration of the attribute. It is considered part of the syntax even
> though it isn't part of the ABNF.
> 
> Of course the semantics of its use is also tied up in whether it is
> session or media level, and so needs to be described as part of the
> semantics as well.

Hi Paul,

I guess my concern is this: now that we've decoupled ICE from SDP, we
might lose synchronization between SDP bindings and non-SDP bindings,
which wouldn't be good for interoperability. As I understand it, the
ice-options attribute can include a variety of tokens (e.g., 'ice2' or
'rtp+ecn'). Should it be the responsibility of a token spec to specify
whether a token is session level or media level, or is that best handled
at the attribute layer? I am not an SDP expert so I might be missing
something fundamental...

Peter