Re: [MMUSIC] AD review: draft-ietf-mmusic-image-attributes-06

[Robert Sparks <[e-mail]>]

inline

On Aug 20, 2010, at 8:07 AM, Kyunghun Jung wrote:

> Hi,
>  
> Let me answer some of your questions.
>  
> Kyunghun Jung
> Samsung Electronics Co., Ltd.
> 
> ------- Original Message -------
> Sender : Robert Sparks<[e-mail]>
> Date : 2010-08-20 03:06 (GMT+09:00)
> Title : [MMUSIC] AD review: draft-ietf-mmusic-image-attributes-06
> 
> (There is a copy of these comments at http://www.nostrum.com/~rjsparks/mmusic-image-attributes.txt 
> for awhile in case any mail clients re-wrap the lines in a way that makes this too hard to read). 
> 
> The document places requirements on "high end" devices without defining them or characterizing 
> them beyond what you can take away from the name. Without a more objective definition, these 
> requirements effectively reduce to "Some implemantation MAY" do these things. If that's what you 
> mean, please say it that way. If you really are trying to inflict SHOULDs and MUSTs on something, 
> please reduce the ambiguity in what they are.
> -> This attribute is initially designed with handsets supporting 4G mobile video telephony in mind
> where media is transported over very expensive bandwidth but its nenefits can be enjoyed by other
> devices if quality loss and unnecessary computation from rsizing is to be avoided.
>  
> Although we at 3GPP consider "high end" as the description of mobile communications devices with
> high computational power with high resolution display (but smaller dot picth), supporting mutiple radio
> frequency bands or roaming capability, we did not find enough reason to introduce such device specifications
> or classifications of a specific inductry or product in the Internet draft.
> 

The problem is that without that definition, you have normative requirements in this draft that are ambiguous.
Saying that a "high end" device MUST do something without characterizing the device does not help
interoperability. If you are not willing to provide something an implementer can use to unambiguously know
the requirement applies to his implementation, you need to change all of these requirements to MAY.

> 
> I'm surprised that there hasn't been more discussion around the message size bloat 
> associated with including an imageattr line with "send * recv *" in offers where the offerrer  
> isn't planning to use this document's extension. Have I missed that searching the archives? 
> If so, please provide a pointer. Were other negotiation mechanisms considered? 
> Since the problem was initially discussed at 3GPP in June 2007 and informed to IETF MMUSIC,
> hundreds of emails containing comments and answers were exchanged. The negotiation mechanisms
> were designed here based on the received suggestions and corrections. In fact, initial idea was to
> recycle (extend the usage of) existing SDP attributes, not defining a new one.
>  
> It might be difficult to zip the hundreds of e-mails exchanged here and send it to you or release in the
> mailing list but let me introduce key documents which might be used to understand the background
> of asking IETF to solve this problems.
> http://www.3gpp.org/ftp/tsg_sa/WG4_CODEC/TSGS4_44/Docs/S4-070408.zip
> http://www.3gpp.org/ftp/tsg_sa/WG4_CODEC/TSGS4_47/Docs/S4-080083.zip
> 
> The document should better capture the motivation for allowing the answerer to create attribute 
> value entries that aren't covered by the offer's proposal. It should better describe _when_ additional 
> offer/answer exchanges would be induced (informing the operator as well as the implementor), and show 
> that this can't result in the infinite ping-pong loop that the general pattern of answering with a 
> subset of the offer was created to avoid in the first place.  
> -> It is not forbidden to add new sizes in the SDP answer and delay from additional negotiation is
> left as the discretion of the implementation. Those intending to use their bit-rate in a more favourable
> way need to take the risk of longer negotiation.
> 
The draft needs to say when this will happen. I think you missed an important part of my comment.
What prevents this "longer" renegotiation from being infinite?

> 
> 
> In section 3.1.1, the document should call out where to find the definition of 
> WSP and DIGIT you are intending to use.  The last two lines of comments at the 
> end of key-value's ABNF need to start with semicolons 
> 
> 3.1.1.2: second bullet (o), last sub-bullet (*): It's not clear what "should not on its own include" 
> means. Are you trying to say that if an offer doesn't include an a=imageattr for a stream, then the 
> offer MUST NOT include one? 
> 
> 3.1.1.2: third bullet (o), first sub-bullet (*): This use of SHOULD is unclear. When would the 
> offer _not_ resort to other methods? I think you're trying to say that if the image attribute 
> is not included in the answer, then the offerer knows the answerer does not support this mechanism 
> and continues to process the answer as if this mechanism had not been offered. 
> 
> 3.1.1.2: second bullet (o), second sub-bullet (*): This has the same issue as the first sub-bullet. 
> Assuming I understand the intent, this would be clearer if you wrote along these lines: 
>   * If the image attribute is included in the SDP answer, but none of the entries are usable or 
>     acceptable, the offer MUST initiate a new offer/answer exchange without using the image attribute. 
> The draft should also indicate what to do with any media packets that might arrive before that second 
> offer/answer completes. 
> 
> 3.2.1: first paragraph - This MUST NOT belongs in 3.1.1.2, (at the last sub-bullet of the second bullet 
> if I've understood you), and it applies to _all_ answers, not just those to initial invites. I suggest 
> changing this sentence to something along these lines: "When the initial offer does not contain the  
> 'imageattr' attribute, the rules in section 3.1.1.2 require the attribute to be absent in the answer." 
> 
> 3.2.6 : It's not clear what the purpose of this section is. If it's only to call out the problem and 
> capture the discussion of some possible solutions (but not specify a solution), that should be stated 
> more clearly. Right now, the text could be read to be recommending the solution in the 3rd bullet of 
> 3.2.6.4, but it's not well specified. Either way, the document should note that using port 0 this way 
> may defeat the return of any information about image attributes from the answerer in the first  
> offer/answer round. Per 3264: 
>    A port number of zero in the offer indicates that the 
>    stream is offered but MUST NOT be used.  This has no useful semantics 
>    in an initial offer, but is allowed for reasons of completeness, 
>    since the answer can contain a zero port indicating a rejected stream 
>    (Section 6).  
> In section 6, it notes that media formats for rejected streams are ignored. Devices wishing to save 
> bandwidth may surpress any attributes associated with that m-line.  Also, be very careful with the  
> semantics - when a slot for a rejected stream is reused, it is adding a new media stream, not  
> modifying/updating an existing stream. 
> 
> 4.2.3: The example claims it is presenting a complete SDP offer - it isn't. It would be more 
> accurate to say "more of the SDP offer is shown" 
> 
> The draft needs to note that the coupling between the attributes in the offer and answer is associated 
> with the codec, not the payload type number. It's possible to negotiate different payload type numbers  
> for the same codec in the pair of unidirectional streams assocated with an m line. The offer can  
> indicate a PT number of 101 and the answer can use PT 102, for example. If that happens, the offerer  
> will be sending RTP with payload frames labeled 102 and the answerer will be sending frames labeled 101. 
> (But the answerer's SDP would have contained a=imageaddr:102). 
> 
> It would also be good to call out what's allowed with this parameter when an m-line has more than one  
> payload type number mapped to the same codec. 
> 
> Nits: 
> 
> Introduction: last paragraph first sentence : This sentence doesn't parse for me. 
> Is it trying to say "This document is limited to point-to-point unicast communication scenarios"? 
> 
> 3.1.1.2: first bullet (o), second sub-bullet (*): Should "recommended" be in upper case? 
> 
> 3.2.9 : first paragraph: The MUST requirement in the 2nd sentence misrepresents what the answer is. 
> There are no unknown parameters to remove from the answer - the answer is not a copy of the offer, 
> it is content that the answerer generates. It would be better to say the answerer MUST ignore 
> any unknown parameters in the offer and MUST NOT include them in the answer it generates. 
> 
> 4.1 : The last part of this example is confusing. Instead of "If Bob does not support any x and y 
> resolution in any of of (sic) the X and Y ranges", perhaps you could say "If Bob does not support 
> any x and y resolution in one of the provided send or recv ranges". Then instead of starting the 
> next sentence as "In this case", start it as "For instance, if Bob doesn't support any of the offered 
> alternatives in the recv attr-list in the offer", and delete the sentence starting "In the above example". 
> 
> _______________________________________________ 
> mmusic mailing list 
> [e-mail] 
> https://www.ietf.org/mailman/listinfo/mmusic 
> 
>