- From: Cyril Concolato <cyril.concolato@enst.fr>
- Date: Tue, 04 Nov 2008 08:51:09 +0100
- To: Doug Schepers <schepers@w3.org>
- CC: www-svg <www-svg@w3.org>
Hi Doug, I am satisfied with the changes you made and I think this was my last comment. I'm done. Thanks. Cyril Doug Schepers a écrit : > Hi, Cyril- > > (Note: Cyril sent this to me offlist, so I'm including www-svg for the > public record.) > > As a high-level comment, I'll just note that you and I have slightly > different interpretations of the SMIL spec, but that you're probably > right. However, as you note, the SMIL spec is clear enough where the > SVG spec is a little vague; as you acknowledge below, this is sufficient > for the SVG 1.2 Tiny spec, but we should revisit and clarify it for SVG > 2.0 Core. We should bring the SYMM WG into the larger discussion. > > Comment inline... > > Cyril Concolato wrote (on 11/3/08 5:35 AM): >> Hi Doug, >> >> Doug Schepers a écrit : >>> Hi, Cyril- >>> >>> Thanks for your comment. >>> >>> Cyril Concolato wrote (on 10/13/08 7:03 AM): >>>> * Section 12.1.1 Media timeline and document timeline >>>> The specification says that media elements which are loaded only when >>>> they become visible for the fist time may introduce delay. In that case, >>>> what document time corresponds to the beginning of the media timeline ? >>>> The begin value or the begin value + the delay ? >>> The latter. >> I would have expected the former to be consistent with the behavior of >> media and 'timelineBegin' which says: >> "In that case, the beginning of the actual playback of the media will be >> delayed, but the media begin time in the document timeline will remain >> as specified. " That is why I thought that the UA had to adjust the >> playback. > > I can see both readings, but I'm not religious about this point. > > >>>> Is it correct to say >>>> that assuming timelines are locked, then the UA has to adjust playback >>>> to cope with the load time (similarly to a timelineBegin=onStart)? >>> When timelines are locked, it is with regards to synchronization of >>> pacing, not a single absolute timeline. See the SMIL spec for more >>> details. [1] >> I don't understand what you mean by "synchronization of pacing". The >> terms 'pace' or 'pacing' are not used in the SMIL spec you reference. It >> is my understanding that when timelines are locked it is as if there was >> a single timeline. See the following quote from the SMIL spec: >> "the syncBehavior attribute controls whether the media element can slip >> while the rest of the document continues to play, or whether the time >> container must also wait until the media delivery catches up." >> >> On this problem, the SVG spec is not wrong, it is just unclear but the >> SMIL spec is actually clear, so I can live with the SVG spec as is. > > Ok, cool. > > >>>> Replace: >>>> "the value of the 'syncBehavior' attribute are replaced from the default >>>> one ('default', interpreted as 'locked' in that case since the >>>> 'syncBehaviorDefault' is not specified) to 'independent'." >>>> with: >>>> "the value of the 'syncBehavior' attribute are set to 'independent'." >>>> The current wording 'interpreted as locked' is ambiguous. The actual >>>> interpretation of 'default' is implementation specific and the spec >>>> seems to imply that it should be interpreted as locked all the time. >>> Yes, you seem to be correct. [2] This would ideally be specified more >>> definitively, but that's up to SMIL. I have removed the offending >>> passage. >> Thank you. > > Yup. > > >>>> Could you clarify what the exact locking behavior is ? If a media >>>> timeline is locked to the document timeline and if the video is paused, >>>> what happens ? Is the document timeline locked? >> I meant here "Is the document timeline paused?", not locked. But I think >> you understood me. > > Yup. > > >>> This is defined in SMIL [1], but the short answer seems to be that >>> pausing the media (or more precisely, the media time container) is >>> independent of the document timeline (which is the master time >>> container). Media that has been paused and then restarted will sync >>> with the pace of the document, not the elapsed time. >> This is not my interpretation of the SMIL spec you point. If you look at >> the 'speech' example. It says that if either the audio or video element >> is paused, then the entire 'speech' (i.e. the entire 'par' time >> container, similar to the 'svg' time container) is paused. >> Again, the SVG Spec is not wrong. I can live with it as is. > > Ok. > > >>>> The xlink:href of audio and video says: >>>> "When the value of the attribute is animated or otherwise modified, if >>>> the media play time can be controlled, then the media timeline is >>>> restarted. " >>>> First, the term 'play time' is not defined. I suggest changing "media >>>> play time" to "media timeline". >>> Yes, thanks, we've changed all instances to match your suggestion. >> Thanks. > > No problem. > > >>>> But if the media timeline is locked, >>>> changing the xlink:href to a new value, without changing the begin >>>> value, will not restart the media timeline (or it will restart it but at >>>> time = the document time when the change happened). What happens if the >>>> media timeline cannot be controlled? >>> I have added wording to address when the media timeline cannot be >>> controlled. >> Here we have 2 main options: >> A) the media time line cannot be controlled (broadcast) >> B) the media time line can be controlled >> For A), there is nothing we can do. If you change the xlink:href from, >> say Broadcast Channel 1 to Broadcast Channel 2, you get the media time >> that's being broadcasted. >> For B), there are 2 sub-options: >> B.1) the media time line is locked to the document time line >> B.2) the media time line is independent from the document time line >> >> In B.1), you can compute the media time as the difference between the >> current document time and the begin time. So changing the xlink:href >> cannot have an effect on the media time and this is a good thing. I >> think we can illustrate this on the problem of keeping synchronization >> upon stream switching >> Example: >> At document time = X, >> <animation syncBehavior="locked" begin="0" xlink:href="subtitle_fr.svg"/> >> At document time = X + 10, you change xlink:href to subtitle_en.svg, >> your internal document is equivalent to: >> <animation syncBehavior="locked" begin="0" xlink:href="subtitle_en.svg"/> >> At document time = X + 20, you revert to subtitle_fr.svg, your internal >> document is equivalent to: >> <animation syncBehavior="locked" begin="0" xlink:href="subtitle_fr.svg"/> >> >> You don't want the subtitle to restart from 0 each time you change the >> xlink:href because this would loose the synchronization. You cannot do >> it via scripting (with only one animation element) because SVG does not >> have an API to query the media time of a timed element. >> In B.2), indeed the media time line needs to be restarted. >> >> Concerning your editing, I don't understand the added wording: >> "If the media timeline cannot be controlled, then it will continue from >> the first available time." >> Given the additional comments above, I think you'd better say: >> "When the value of this attribute is animated or otherwise modified, if >> the media timeline can be controlled, then the media timeline is >> restarted *only if the syncBehavior attribute is set to 'independent'*. >> If the media timeline cannot be controlled, then the media timeline is >> unaffected by such modification." > > That is a sensible argument, and I agree that your wording is better > than mine. I have made this change, as requested. > > >>>> The specification should say what happens when a video/audio element >>>> points to a file or a set of streams containing multiple audio tracks >>>> (english and french, or AAC and AC3) or multiple video tracks. >>> That's a good suggestion, and if this is not already covered by SMIL, we >>> will define this behavior in SVG 2.0 Core. For now, this will have to >>> be implementation-specific, and we will look at implementations to >>> inform how we specify it later. >> That's fine. > > Cool. > > Since you've stated that you can live with the spec as it is, modulo the > change you've requested above (which I've made), I am going to close > this issue, and indicate your satisfaction in our Disposition of > Comments. Please let us know if this doesn't work for you. > > Note that I agree we should reopen this issue for further clarification > in the near future. Thanks for all your insightful analysis. > > Regards- > -Doug > -- Cyril Concolato Maître de Conférences/Associate Professor Groupe Mutimedia/Multimedia Group Département Traitement du Signal et Images /Dept. Signal and Image Processing Ecole Nationale Supérieure des Télécommunications 46 rue Barrault 75 013 Paris, France http://tsi.enst.fr/~concolat
Received on Tuesday, 4 November 2008 07:52:09 UTC