Re: [MMUSIC] draft-ietf-mmusic-sctp-sdp-07 SCTP SDP syntax question

[Christian Groves <[e-mail]>]

Hello Christer,

On 2/10/2014 7:39 PM, Christer Holmberg wrote:
> Hi,
>
>>> I'm happy to use to m-lines. I think it would be good to have
>>> something in the draft describing the usage of multiple DTLS/SCTP
>>> m-lines (with the same port) so that its clear that's the intention.
>> The thing is - it would be bad to give an example using bundle while there is no definition of how bundle works for that. (It is not included in the bundle spec.)
>>
>> That would mean adding the interaction with bundle into this draft, or in a new draft. While that is possible, do we want to take that on now?
>>
>> This would be a good time for Christer to chime in. :-)
> As Paul said, BUNDLE does not specify the usage of multiple DTLS m- lines within a BUNDLE group, and explicitly says that such usage would have to be defined in another specification.
>
> Personally, I'd be quite surprised if people started to transport BFCP, MSRP etc over DTLS/SCTP. People haven't used SCTP for that in the past, so why would they suddenly start now?
[CNG] It seems now people are being forced to use to SCTP due to 
interoperation with WebRTC. They may want to run over a secure 
association etc. There will be new protocols that will be defined.
>
> I think DTLS/SCTP is going to be used mainly for data channel usage, and one can then define data channels to transport whatever (including BFCP and MSRP).
[CNG] Mainly or only? If its mainly then there's a possibility to run 
something else in another SCTP association. If this "something else" 
runs in across a separate SCTP association on a seperate DTLS port then 
its no issue. If this "something else" is a separate SCTP association on 
the same DTLS port then it is an issue. There has been a suggestion to 
use bundle, but bundle doesn't cover this. I think the draft should 
offer some guidance or at least highlight the issues with respect to this.

With regards to the data channels if I do as you say and use 
datachannels for bfcp and msrp in the one DTLS/SCTP association. How 
would I go about signalling that via SDP? Do I use the same method as 
CLUE and define a group for each of these? Do I use ejzak? Do I use 
both? CLUE decided that it didn't need to consider other protocols. 
However shouldn't the scope of the mmusic work be to look at the 
scenarios where you could have multiple protocols?

Regards, Christian

>
> Regards,
>
> Christer
>
>
>
>
>> On 1/10/2014 10:45 PM, Christer Holmberg wrote:
>>> My mistake: I meant SCTP association, which then can contains
>>> multiple SCTP streams.
>>>
>>> Sorry for the confusion.
>>>
>>> Regards,
>>>
>>> Christer
>>>
>>> -----Original Message-----
>>> From: Schwarz, Albrecht (Albrecht)
>>> [mailto:[e-mail]]
>>> Sent: 1. lokakuuta 2014 15:43
>>> To: Christer Holmberg; Christian Groves; Salvatore Loreto
>>> Cc: <[e-mail]>
>>> Subject: RE: [MMUSIC] draft-ietf-mmusic-sctp-sdp-07 SCTP SDP syntax
>>> question
>>>
>>>> I think we need to be careful with the terminology.
>>>> As far as I understand, the m- lines describes an SCTP connection -
>>>> not an SCTP association - and the sctp-fmt defines the usage of that
>>>> SCTP connection.
>>> Christer,
>>> there isn't any concept of an "SCTP connection" (see RFC 4960)!
>>> Rather SCTP association, SCTP path, SCTP stream.
>>>
>>> If you are refering implicitly to the OSI concept of a (N)-connection
>>> (ITU-T X.200), then an OSI (SCTP)-connection would be mapped to the
>>> IETF SCTP association in my understanding.
>>>
>>> Regards,
>>> Albrecht
>>>
>>> PS
>>> Perhaps you mean sth like, ".. the m- lines describes an IP transport
>>> connection for SCTP, which relates to an SCTP association ..."
>>>
>>>
>>> -----Original Message-----
>>> From: mmusic [mailto:[e-mail]] On Behalf Of Christer
>>> Holmberg
>>> Sent: Mittwoch, 1. Oktober 2014 13:43
>>> To: Christian Groves; Salvatore Loreto
>>> Cc: <[e-mail]>
>>> Subject: Re: [MMUSIC] draft-ietf-mmusic-sctp-sdp-07 SCTP SDP syntax
>>> question
>>>
>>> Hi,
>>>
>>> I think we need to be careful with the terminology.
>>>
>>> As far as I understand, the m- lines describes an SCTP connection -
>>> not an SCTP association - and the sctp-fmt defines the usage of that
>>> SCTP connection.
>>>
>>> Depending on the usage, there may then be multiple SCTP associations
>>> - for example as in the case of a data-channel connection usage,
>>> where the number of SCTP associations will depend on the number of
>>> data channels.
>>>
>>> So, if you want to use BFCP and MSRP, my understanding is that you
>>> would need two m- lines (unless, of course, you transport BFCP and
>>> MSRP within data channels).
>>>
>>> Regards,
>>>
>>> Christer
>>>
>>> -----Original Message-----
>>> From: mmusic [mailto:[e-mail]] On Behalf Of Christian
>>> Groves
>>> Sent: 1. lokakuuta 2014 10:45
>>> To: Salvatore Loreto
>>> Cc: <[e-mail]>
>>> Subject: Re: [MMUSIC] draft-ietf-mmusic-sctp-sdp-07 SCTP SDP syntax
>>> question
>>>
>>> Hello Salvatore,
>>>
>>> Thanks for the replies. Please see some comments below.
>>>
>>> Regards, Christian
>>>
>>> On 1/10/2014 5:29 PM, Salvatore Loreto wrote:
>>>> Hi Christian,
>>>>
>>>> thanks a lot for reading and reviewing the draft see more in line
>>>>
>>>> On Sep 30, 2014, at 3:48 PM, Christian Groves
>>>> <[e-mail]> wrote:
>>>>
>>>>> I seemed to have got cut and paste happy in 1) below. The "1*(SP
>>>>> sctp-fmt)" should only apply to "DTLS/SCTP".
>>>>>
>>>>> Christian
>>>>>
>>>>> On 30/09/2014 10:12 PM, Christian Groves wrote:
>>>>>> Hello Salvatore,
>>>>>>
>>>>>> There seems to be several issues with the updated syntax:
>>>>>>
>>>>>> 1) Cl.4.1 Media Description: With the updated syntax the ability
>>>>>> to have more than one usage of the SCTP association has been lost
>>>>>> as "sctp-fmt" is a single value.
>>>> <Sal>
>>>> there have been several discussion about the ability to have more
>>>> than one usage of the SCTP association, and has been agreed not to allow it.
>>>> So each 'm' line describes a single SCTP association;  each
>>>> association has only one single specify usage.
>>>> </Sal>
>>> [CNG] So how would I specify in SDP a scenario where BFCP uses a
>>> DTLS/SCTP association and MSRP uses a DTLS/SCTP association where
>>> they both use the same DTLS connection? Are these to be separate m-lines?
>>>> ..snip..
>>>>>> 3) Cl.4.1.2: There appears to be a formal space (SP) missing from
>>>>>> the syntax between association-usage and max-message-size.
>>>>>>
>>>>>> sctpmap-attr = "a=fmtp:" association-usage [max-message-size]
>>>>>>
>>>>>> should be:
>>>>>> sctpmap-attr = "a=fmtp:" association-usage [SP max-message-size]
>>>>>>
>>>>>> The current syntax allows an fmtp usage with no max-message-size
>>>>>> parameter, e.g. "a=fmtp:bfcp". Should this be allowed?
>>>> <Sal>
>>>> this has been discussed back in April or May. It has been decided
>>>> that the max-message-size parameter is optional.
>>>> In the case it is not present the default value is 64K </Sal>
>>> [CNG] I agree that the max-message-size parameter is optional, in
>>> that case an endpoint wouldn't specify the whole a=ftmp: line. It
>>> seems strange to allow the endpoint to specify half a line?
>>>
>>>>>> 5) Cl.4.1.2: "If the parameter is not present, the implementation
>>>>>> should provide a default, with a suggested value of 64K." Is this
>>>>>> 64,000 or is 65,535 bytes?
>>> [CNG] I think you skipped this?
>>>
>>>>>> 7) Cl.10: How will this registry relate to the "WebSocket
>>>>>> Subprotocol Name Registry"
>>>>>> (http://www.iana.org/assignments/websocket/websocket.xml#subprotocol-name)?
>>>>>>
>>>> <Sal>
>>>> It is not related. The 'association-usage' specifies how the SCTP
>>>> association will be used.
>>>> For example in the case of webrtc-datachannel value it would
>>>> indicate how to pair certain streams, how to consider specific stream etc.
>>>> as described in
>>>> https://tools.ietf.org/html/draft-ietf-rtcweb-data-protocol-08 and
>>>> in http://tools.ietf.org/wg/rtcweb/draft-ietf-rtcweb-data-channel/
>>>>
>>>> the WebSocket Subprotocol Name Registry is used to populate the
>>>> protocol field in the protocol field in
>>>> https://tools.ietf.org/html/draft-ietf-rtcweb-data-protocol-08#secti
>>>> on
>>>> -5
>>>>
>>>> </Sal>
>>> [CNG] So if I use the SCTP association for "bfcp" then i'd need to
>>> document this in the new registry associated with the draft? and if I
>>> then used "bfcp" for webrtc-datachannel then i'd need to update the
>>> websocket registry. Two places for the same protocol?
>>>
>>>
>>> _______________________________________________
>>> mmusic mailing list
>>> [e-mail]
>>> https://www.ietf.org/mailman/listinfo/mmusic
>>>
>>> _______________________________________________
>>> mmusic mailing list
>>> [e-mail]
>>> https://www.ietf.org/mailman/listinfo/mmusic
>>>
>> _______________________________________________
>> mmusic mailing list
>> [e-mail]
>> https://www.ietf.org/mailman/listinfo/mmusic
>>
> _______________________________________________
> mmusic mailing list
> [e-mail]
> https://www.ietf.org/mailman/listinfo/mmusic
>
> _______________________________________________
> mmusic mailing list
> [e-mail]
> https://www.ietf.org/mailman/listinfo/mmusic
>