Re: [MMUSIC] Draft new: draft-holmberg-mmusic-t140-usage-data-channel

[Christer Holmberg <[e-mail]>]

I have created a pull request, which will be used for the changes based on Gunnar's comments:

https://github.com/cdh4u/draft-datachannel-t140/pull/5

Regards,

Christer

On 22/08/2019, 13.39, "mmusic on behalf of Christer Holmberg" <[e-mail] on behalf of [e-mail]> wrote:

    Hi Gunnar,
    
    Thanks you for your support (I assume :) and comments on the draft! 
    
    See inline.
      
    >A couple of comments:
    >1) In 3.2, the attribute "cps" is misspelled "cpc" once.
    
    Will fix.
    
    ---
    
    >2) Section 5 has some historical references to real-time text transports that may not be of much interest anymore 
    >and instead confuse the reader, while some other more relevant transports may be added. 
    
    I took these from the schwarz draft. You probably know better than me which ones are relevant, so feel to suggest which one(s) should be removed, and which one(s) should be be added :)
    
    >I would also like to discuss if it could be possible to have a few general recommendations on the webrtc to sip/rfc4103 case without 
    >the problems you see with having a detailed gateway section.
    
    The second last paragraph covers some things on the media plane (out of order and loss of RTP packets) that I think are worth mentioning.
    
    As far as SDP interworking is concerned, this draft defines the m- line for T.140 data channel, and RFC 4103 defines the m- line for T.140 RTP, and the interworking should be very straight forwards. Do you have something specific in mind regarding general recommendations?
    
    ---
    
    > 3) Reliability. Section 3.1 implies that the channel is used in the reliable and ordered mode. We have been discussing back and forth
    > if that is the right choice for real-time text. I tend to think it is, but it might be useful to discuss it once again. The traditional user 
    > requirement on real-time text is that produced characters shall be presented to the receiver within one second from their creation. 
    > Modern usage in speech-to-text applications may require more rapid transmission. As I understand it, the reliable mode of the 
    > data channel may imply long periods of choked transmission in case of network problems or by influence of heavy transmission 
    > in another channel. As long as this happens only in case of network problems, I now tend to think that that might be acceptable. 
    > The effects of being forced to use an unreliable channel are so far-going so I would like to avoid that.
    > However, the word "reliable" is misleading. A "reliable" channel is not really reliable. It can break in case of problems. 
    
    True, but "reliable" is the terminology used in both RFC 4960 (SCTP) and draft-ietf-rtcweb-data-channel.
    
    > I think some recommendations should be inserted in section 4 about what to do when a channel breaks. The natural action 
    > would be for both sides to try to figure out what was the last T.140 data that was transmitted and received, and then try to 
    > reconnect and resume transmission if successful. If any T.140 data was lost during the break, that state should be marked 
    > by inserting the "missing data" T.140 indicator in the received stream. There needs of course also be a recommended action 
    > if it turns out to be impossible to reconnect after a low number of retries. 
    
    I can for sure add some text about that. Are there generic T.140 recommendations for failure that we can reference, or do you think there is something T.140 data channel specific? 
    
    Regards,
    
    Christer
    
    
    _______________________________________________
    mmusic mailing list
    [e-mail]
    https://www.ietf.org/mailman/listinfo/mmusic