I completed another review of the draft-ietf-taps-transports-03 in the context 
of a "gap analysis” when thinking of  multisource/multidestination (MS/MD) 
video sessions. What I mean by MS/MD is:
a) multisource: for example real-time and progressive downloads combined with 
social media and user generated content (aka social television)
b) multidestination: co-vieing of content either in real-time or via 
progressive downloads (binging on Netflix)

The draft is good and is supplying I think a common way of evaluating the 
different transport protocols. I want to highlight just a few items that i 
think are still missing 1.  more details on secure transports and 2. a good 
discussion of what reliable transport means (in the sense of RMT and more). See 
details below.

1. Secure transports:
I think the inclusion of (D)TLS will be good as it will highlight the needs for 
secure transports to ensure some privacy. Also discussing TCP-AO, TCP-INC and 
IPSEC would make it complete. However in the context of MS/MD I think we should 
make sure we look into the delays that these protocols incur. I guess we can 
make the assumption that all parties in the MS/MD will use the same transport 
so would all experience simlar encryption delays. I remember that at the 
beginning of the WG we had made no such assumption on the use of uniform 
transports across sessions. Maybe we need a discussion? List what do you think?

2. Reliable transport:
I have a issue with TCP being put as completely reliable; I understand why it 
is there but in terms of the MS/MD application TCP is not "fully reliable" it's 
actually not reliable especially when it comes to latency. I think the text 
could be enhanced by the mention of this.

Video stops and sync is lost between the streams (in my world we call that 
"buffering events") and the quality of experience of viewing groups will be 
greatly impacted. For realtime I suppose we can look into things like WebRTC 
which I do not think is being directed at things other than voice/video 
conferencing right now but of course could be extended. This also implies the 
use of a browser which is not always the case. But for progressive downloads 
which is most of the consumed video and for the  content to keeps in sync and 
relevant (when one stream stops the others become out of scope) there needs to 
be more reliability. Of course the FEC protocols developed for RMT are still 
valid even if we look at multiple unicast sessions and the document could refer 
to that? I also think that non-block approaches (newer sliding window codes) 
could be considered. And of course we should mention how congestion control 
fits in all of this :-)

Finally Table 4.1 has a number of features that are being addressed by existing 
protocols which is fine. Do we want have a paragraph on features that are 
needed but could be supplied by either not fully standardized protocols or 
protocols developed in the future (see reliability above)? 

But in any case great job and looking forward to discuss it in Dallas.

Marie-José

Marie-Jose Montpetit, Ph.D.
[email protected]
@SocialTVMIT

Attachment: smime.p7s
Description: S/MIME cryptographic signature

_______________________________________________
Taps mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/taps

Reply via email to