Re: [MMUSIC] Review of draft-ietf-mmusic-sctp-sdp-22

[Christer Holmberg <[e-mail]>]

Hi,

I¹ve created a pull request with the changes based on Brian¹s comments.

https://github.com/cdh4u/draft-sctp-sdp/pull/10


Regards,

Christer


On 05/02/17 21:05, "Brian E Carpenter" <[e-mail]> wrote:

>Hi Christer,
>On 06/02/2017 03:14, Christer Holmberg wrote:
>> Hi Brian,
>> 
>> Thanks for your review! Please see inline.
>> 
>> ...
>> 
>> Comments:
>> ---------
>> 
>>> Two points I noted in the writeup:
>>> "There are existing implementations of earlier versions of the
>>>document..."
>>>
>>> Excellent, but I wonder why we don't see Implementation Status
>>>sections under RFC 6982 in more Last Call drafts.
>> 
>> Perhaps a topic for an IETF chairs lunch presentation?
>
>Or at least a short reminder!
>
>> 
>> I have to admit that I did not know about 6982 (now obsoleted by 7942).
>>My suggestion would to not collect that information for this draft at
>>this point of time, as it could take some time.
>
>Agreed.
>
>>> "IPv6 address examples are not necessary since IP version differences
>>>are immaterial to the purpose of the specification."
>>>
>>> It's just as easy to give an IPv6 example though, and more future
>>>proof.
>> 
>> I can modify the address in the example, from IPv4 to v6.
>
>As you wish.
>
>> Minor issue: (almost a nit)
>> ------------
>> 
>> 1.  Introduction
>> ...
>>>>   NOTE: Due to the characteristics of TCP, usage of 'TCP/DTLS/SCTP'
>>>>   will always force ordered and reliable delivery of the SCTP
>>>>   packets, which limits the usage of the SCTP options.  Therefore, it
>>>>is
>>>>   RECOMMENDED that TCP is only used in situations where UDP traffic
>>>>   is blocked.
>>>
>>> Why would one choose 'TCP/DTLS/SCTP' rather than just 'TCP/TLS'? I
>>>don't object to it being specified, but since you
>>> don't support multihoming or multiple associations, what is the use
>>>case, in a few words?
>> 
>> SCTP supports multiple SCTP streams over a single association, and you
>>can still use that with 'TCP/DTLS/SCTP'. E.g, the WebRTC Data Channel
>>protocol uses two streams to realize a data channel.
>
>OK.
>
>(I now remember that I raised essentially the same question when reviewing
>draft-ietf-clue-datachannel. Actually I'm glad to see SCTP proving
>useful.)
>
>> I could modify the first sentence as following:
>> 
>>         "NOTE: Due to the characteristics of TCP, while multiple SCTP
>>streams
>>          can still be used, usage of 'TCP/DTLS/SCTP' will always
>>force...."
>
>OK
>
>> 
>> Nits:
>> -----
>> 
>>>> 4.4.2.  SDP Media Description values
>>>>
>>>>      m= line parameter        parameter value(s)
>>>>     
>>>> ------------------------------------------------------------------
>>>>      <media>:                 "application"
>>>>      <proto>:                 "UDP/DTLS/SCTP" or "TCP/DTLS/SCTP"
>>>>      <port>:                  UDP port number (for "UDP/DTLS/SCTP")
>>>>                               TCP port number (for ""UDP/DTLS/SCTP")
>>>
>>> I think the last line should be: TCP port number (for "TCP/DTLS/SCTP")
>> 
>> Correct. Will be fixed.
>> 
>>> There is some inconsistency in the use of quotation marks:
>>>"UDP/DTLS/SCTP" or 'UDP/DTLS/SCTP'
>> 
>> I'll fix that. I will use single quotes.
>
>Thanks,
>    Brian