Re: [clue] Gen-ART review of draft-ietf-clue-telepresence-requirements-06.txt

["Duckworth, Mark" <[e-mail]>]

Comments inline.
Mark Duckworth

From: clue [mailto:[e-mail]] On Behalf Of Stephen Botzko
Sent: Wednesday, November 27, 2013 5:28 PM
To: Mary Barnes
Cc: [e-mail]; [e-mail]
Subject: Re: [clue] Gen-ART review of draft-ietf-clue-telepresence-requirements-06.txt

Dan - thanks so much for reviewing this.

Comments are in-line

Stephen Botzko

On Thu, Nov 21, 2013 at 4:39 PM, Mary Barnes <[e-mail]<mailto:[e-mail]>> wrote:
On Thu, Nov 21, 2013 at 10:38 AM, Paul Kyzivat <[e-mail]<mailto:[e-mail]>> wrote:
Authors,

What do you think of these comments?
If you aren't in agreement with all of them, then we should have a dialog with Dan about them.

If in agreement, then lets discuss what changes are appropriate.

More inline.


On 11/21/13 7:08 AM, Romascanu, Dan (Dan) wrote:
I am the assigned Gen-ART reviewer for this draft. For background on Gen-ART, please see the FAQ at

<http://wiki.tools.ietf.org/area/gen/trac/wiki/GenArtfaq>.

Please resolve these comments along with any other Last Call comments you may receive.

Document: draft-ietf-clue-telepresence-requirements-06.txt
Reviewer: Dan Romascanu
Review Date: 11/21/13
IETF LC End Date: 11/27/13
IESG Telechat date: (if known)

Summary: It's a good and well written document, almost ready, but a number of issues should be clarified before approval.

Major issues:

1. The definitions of Left and Right in Section 3 use an entry article in Wikipedia ('Stage Directions'). I am both uncomfortable with the usage of Wikipedia articles (subject to change, subject to attacks) as references, and I also believe that the current definitions (@11/21/13, US morning hours) are ambiguous. Actually the Wikipedia article does not include definitions for Left and Right, and speaks about 'house left' (equal to 'stage right') and 'house right' (equal to 'stage left'). If another specification in CLUE will use the terms defined here it is not clear which one is meant.

This seems like a reasonable point to me. Is there a more stable reference we can use? If not, can we just define these terms in the document?
[MB] I'm almost of a mind to just remove those definitions.  The consensus was that the terms should be qualified when they are used and the framework does an excellent job of defining more precise terms that properly qualify "left" and "right" and differentiate from stage right, etc. as appropriate.  We use "left" and "right" in one place in this requirements document and I think the use is general enough that we don't need a definition (in this document).   The wikipedia reference has been in the document since day 1 as a WG document (it was added in -03 of the individual document).   The discussions around "left"/"right" on the list were in the context of the definitions document and again, consensus was that the terms should be qualified (as they are in the framework), thus I don't think the current definitions serve any value. [/MB]
[SB] I agree with Mary that these definitions can simply be removed.
[/SB]
[Duckworth, Mark] I agree too.

2. In REQMT-1d it is not clear what 'the extent of individual video captures' means

3. In REQMT-2d it is not clear what 'the extent of individual audio captures' means

We have had our own concerns about this. I think we want to say *something* but its unclear to me exactly what. Perhaps we could focus on what we want to use this for, and explicitly leave details of what it is for the FW.
[MB] Can you point to the threads of discussion where you feel this concern with this phrasing has been raised?   There were certainly concerns raised about those requirements (i.e., 3D), but I'm not reading this as Dan's concern.   My understanding is that the word "extent" is used in the context of video imagery from the perspective of the camera - i.e., the "angular extent" or something like that.  I think we should replace the word "extent" with something more specific.  Anyone have a suggestion?
[SB]
The intent of REQMT-1d was to say that CLUE needs to identify the area that each camera captured in the room - so it would be clear how views from different camera related to each other, and also if there were any gaps, etc.  Identifying the scale of the view is related, but is captured concisely in requirement 6.  Proper rendering in a multi-camera telepresence system requires this knowledge - today it is generally hard-wired into the room design.  But that doesn't result in an interoperable experience, and it needs to be signaled if you want to get appropriate rendering between rooms with disparate designs.

For audio [REQMT-2d], there is an analogous need to identify the area of coverage of a microphone.  The need to correlate the audio and the video is related, but is captured separately in REQMT-3.

The discussions on the list related to audio and video area of coverage were about whether the model in the framework was adequate.  Personally I believe it is.  But what really matters here is whether we are clearly stating the requirement, not the adequacy of the proposed specifications.
One alternative term is "area of coverage", which is commonly used in surveillance applications to describe a similar concept.  If people think that might be confused with radio coverage (e.g. an LTE coverage map) we could draft a definition and add it to section 3.
[/SB]
[Duckworth, Mark] I agree with Steve about the intent.  I agree "area of coverage" is a better term than "extent" for this purpose.


And, note these requirements were added based on discussion at IETF-82:
http://www.ietf.org/proceedings/82/slides/clue-1.pdf

 [/MB]
        Thanks,
        Paul

4. REQMT-10: The solution MUST make it possible for endpoints without
               support for telepresence extensions to participate in a
               telepresence session with those that do.

More clarity is needed to explain what level of participation in a session (and what limitations) are expected from endpoints that do not support the telepresence extensions. I assume there are limitations, otherwise extensions would not be needed.
[SB] The intent was to simply state that the solution must not prevent direct interoperability with SIP devices (either phone or videoconferencing) - without the need for a middlebox to convert signaling (as would be required with rtcweb for instance).   The limitation is simply that telepresence extensions can't be used.  I do not see a need to describe the limitations of existing SIP systems in this requirements document.

Suggested text that makes the requirement more clear would be helpful.  This is a bit tricky to word, since SIP itself has no requirement that says SIP endpoints must interoperate with each other w/o middleboxes to assist. And in fact they often don't.  We could alternatively drop the requirement.
[/SB]
[Duckworth, Mark] I agree with Steve about the intent.  How about something like "The solution must make it possible for devices that support CLUE to interoperate directly with SIP devices that do not support CLUE" ?

5. REQMT-15 (last bullet): I could not really figure out what is meant here

*  There can be variation in placement, number and size of
                  presentations
[SB]At this point I am not certain either.  We might have meant that different use cases had different placements, number, and size of presentations. Or that existing telepresence systems handle presentations in different ways.

It is not worded as a clear requirement, and it says "can", not "should" or "must".  My recommendation is to delete that bullet.
[/SB]
[Duckworth, Mark] I think we meant this:

             - There can be multiple simultaneous presentation media streams

            - Presentation media streams can be spatially related to each other


6. REQMT-16:  The solution MUST include extensibility mechanisms.

What 'extensibility' is meant here? Extending the telepresence sessions, or extending the specifications for new functionality in the future?
[SB] What is meant is extending the specifications for new functionality in the future.  I suggest that we re-word to "New protocols developed for the solution MUST include extensibility mechanisms"
[/SB]
[Duckworth, Mark] Steve's suggestion is more clear.


Minor issues:

1. If this is the first and basic document to be read by somebody who starts to browse through the CLUE specifications, it would be good to define or expand some place in the introduction what CLUE is.
[SB] I am not sure if the CLUE WG has a recommended order for reading the various documents.  Personally I would recommend that people start with the use cases.  In any event, I think the document purpose is properly captured in the abstract - it is not intended to be an introduction to telepresence generally or as an orientation to CLUE specifications.
[/SB]
[Duckworth, Mark] I agree.

2. In the Introduction Section I see:
  These
    requirements are for the specification, they are not requirements on
    the telepresence systems implementing the solution/protocol that will
    be specified.

But then all requirements in section 5 start by 'The solution ...' which is ambiguous. It would be good to mention (again) in section 5 that by what is meant by the 'solution' is the definition of the protocol, not the implementation.
[SB] I agree.  We should also use "specifications", since current plans call for more than one standards-track RFC
[/SB]
[Duckworth, Mark] I agree

3. I will defer to the Security Review to comment whether REQMT-18 is enough from a security point of view - especially as the information about media captures may include explicitly or implicitly details about equipment on each site, etc. To me it seems a little bit too vague.
[SB] I don't see it as vague - it pretty clearly says that only the mechanism needs to be secure.  I think you are saying that the security section might be incomplete.

Current protocol designs do not include explicit details about equipment, and I see nothing in the requirements that suggest otherwise.  Some requirements do require transport of information that can be used infer room dimensions (at least of the visible portion of the room) and the audio/video coverage within it.  Can you identify threats that might use that information?
Of course each specification includes its own Security Considerations. Obviously we will address any security review comments in this one.
[/SB].


Regards,

Dan