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

Hi,

> I would prefer disaggregated media to be handled separately and keep ice-options a session level attribute.. Dealing with disaggregated media is a much bigger issue then just adjusting ice-options.
> For instance, why ice-options are handled there and ice-lite is excluded? I think ice-lite is potentially as big of the problem when aggregating media from several sources. Also, I we sure that making singe ICE agent
> dealing with two independent ICE agents does not have any issues? I do not think this was examined in sufficient detail.

If you are going to use disaggregated media with ICE, I think there are more issues than signalling.

For example, keep in mind that the check lists impact each other (e.g., a state change of a pair in list X might impact the state of a pair in list Y). How would that work if check lists X and Y are handled by different ICE agents?

With ICE trickle, one agent would need to know that candidates the other agent(s) is trickling.

Etc.

There would either need to be some "inter-agent communication” mechanism, or we would need to "de-couple” how media streams (check lists) depend on each other.

So, if people want to specify disaggregated-media-with-ICE, I think it should be done as a separate deliverable (possibly in a separate WG).

Regards,

Christer

On Mon, Apr 23, 2018 at 12:45 PM, Marc Petit-Huguenin <marc@petit-huguenin.org<mailto:marc@petit-huguenin.org>> wrote:
This was done as a consequence of merging this draft in ice-bis:

https://www.ietf.org/archive/id/draft-petithuguenin-mmusic-ice-attributes-level-04.txt

There is probably discussions that can be found in the minutes of IETF meetings around that time when the WG agreed to that modification.

On 04/20/2018 11:09 PM, Suhas Nandakumar wrote:
> I am fine with considering ice-options has session level too. I wonder if
> Marc can chime in on considering it as  media level in the ice-sip-sdp.
>
> On Fri, Apr 20, 2018 at 12:07 PM, Peter Saint-Andre <stpeter@mozilla.com<mailto:stpeter@mozilla.com>>
> wrote:
>
>> On 4/20/18 12:37 PM, Roman Shpount wrote:
>>> Hi All,
>>>
>>> I do not think ice-options was ever used as a media level attribute.
>>> Only two ice-options defined in IANA registry
>>> (https://www.iana.org/assignments/ice/ice.xhtml) are ice2 and  rtp+ecn,
>>> which are both session level.
>>>
>>> Furthermore, RFC 5245 section 15.5
>>> (https://tools.ietf.org/html/rfc5245#section-15.5,) defined that
>>> ice-options attribute is session level only. Thus, ice-options should be
>>> kept session level for backwards compatibility.
>>>
>>> If something more granular then session is needed, I think it should go
>>> into candidate extension attribute, not into media level ice-options.
>>>
>>> Bottom line, I think ice-options should be a a session level only
>> attribute.
>>
>> Agreed.
>>
>> In the Trickle ICE spec, we also limited the `trickle` option to session
>> level only.
>>
>> Peter
>>
>>

--
Marc Petit-Huguenin
Email: marc@petit-huguenin.org<mailto:marc@petit-huguenin.org>
Blog: https://marc.petit-huguenin.org
Profile: https://www.linkedin.com/in/petithug

_______________________________________________
mmusic mailing list
mmusic@ietf.org<mailto:mmusic@ietf.org>
https://www.ietf.org/mailman/listinfo/mmusic