Meeting minutes
This meeting
Nigel: DAPT, ttml:role issues, TPAC. Anything to add?
(nothing)
DAPT
Add section about mapping from TTML to the DAPT data model w3c/dapt#216
github: https://
Nigel: I made some changes after discussion last week
… I'm 90% of the way there, needs a re-review
Cyril: I haven't had a chance to read again
Nigel: This is the biggest thing holding up CR
… I wanted to defer defining something an explicit signal for when a div represents a script event, and leave it to another PR
Cyril: I want to re-read, but let's assume we can do that
Nigel: I tried to make the wording about content profiles generic, so there's now a content type
… So it allows us to add other content profiles later, if we want to
… This needs some review
Cyril: Where do we stand on Pierre's question from last time?
… He said in his experience, trying to leave extensibility hooks for a future version wasn't necessarily the best approach,
… as opposed to having a v2 being backwards compatible.
Nigel: I think how it should be read now is that we've allowed ourselves to make that decision in the future
… We don't make any promises that all future DAPT versions can be read by v1 processors
Cyril: I see that it requires compatibility with the declared content profile
… So what about using divs in that context? Do we need provision for grouping of divs in v1?
Nigel: We don't prohibit nested divs, we just define rules from when they're there, if you're going from TTML to DAPT data model
Cyril: So if you receive a TTML2 document with DAPT content profiles in it, if the impl wants to go back to a DAPT content model, you scan the document and identify what's there, ignoring the rest?
Nigel: Yes
Cyril: I'll read again with this model in mind
Cyril: Some other questions. We're also clarifying processor behaviour when there's additional vocabulary
… There's foreign vocabulary and unrecognised vocabulary
… Unrecognised means the impl knows it doesn't support, whereas foreign is any namespace outside the ones defined in the spec
Nigel: Yes, and specific attribute names for those in the global namespace
Cyril: Can be both foreign and recognised or not recognised?
… So they're orthogonal concepts?
Nigel: Foreign is a subset of unrecognised, the way I've defined it
Cyril: So you can split unrecognised into two sets: things inside the spec, and things outside
Nigel: Please leave feedback, I've defined a term in relation to something that's undefined. That needs changing
Cyril: About the behaviour, we have 3 classes of processors: presentation, transformation, validation
… They may or may not behave differently with foreign vocabulary
Nigel: Transformation processors have to prune. Validation is a subset of Transformation
Cyril: Pruning everything except what's in a metadata element or a descendent of metadata element
Nigel: Yes
… And shouldn't put anything in the metadata element that depends on the semantic of the content
Cyril: And presentation processors?
Nigel: They should recognise elements and attributes they don't support
Cyril: So if you know it's a feature you support but there's no profile declaring the feature...
Nigel: That's in Profile Signalling
Cyril: There was a sentence that talked about this: along the lines of you may do something or fix it?
… The second paragraph in the new 6.2
Nigel: There's a question about what the last clause there means
Cyril: May need rephrasing depending on the class of processor
Nigel: I'm resistant to introducing processor classes, it might introduce complexity
… That is something we should change, in section 6.2, second paragraph isn't about document conformance claims, it's about processor behaviour
… Also we should check where the content profiles have been made generic, haven't done it for processor profiles.
Cyril: Is there anything else that this PR addresses?
Nigel: The dark mode change due to ReSpec and W3C stylesheet change
<MattS> Apols - I need to head to another meeting...
<atsushi> https://
Nigel: Worth checking if it works, including the diagrams
SUMMARY: Review to continue
TTML
Permit ttm:role attribute in ttm:desc elements w3c/ttml2#1247
github: https://
Nigel: We last discussed in May. As DAPT defines desc type, so I suggest closing this with no changes, so not at ttm:role on metadata descendents
Cyril: Agree
SUMMARY: Close with no change
TPAC 2024
Nigel: We have joint meeting with APA and MEIG on Monday. Joint meeting Friday with Audio CG and TTWG meetings
Chris: Is there overlap with MEIG in first session Monday?
Nigel: Yes, It's not supposed to be
Chris: Maybe we can combine APA/TTWG and APA/MEIG into one session
… Also Media WG overlaps with MEIG/APA
Nigel: Yes, that worked well last time
Nigel: Chris, let's follow up with the team
Chris: Should we talk about the MSE and substitles issue you raised in the MEIG repo?
Nigel: Currently subtitles are all out of band with MSE. Not clear why that is. In the past Mark Watson thought it was a good idea
… Could be worth revisiting, now we have more experience, would it solve problems?
Chris: To summarise this would be TTML or IMSC in MP4 , processed through MSE and surfaced into the TextTrack API
Nigel: Yes. Also WebVTT. The CMAF profile requires IMSC in MP4
Chris: Include in the MEIG/TTWG joint meeting?
Nigel: For that meeting, we also want to cover DAPT
Nigel: Another use case, is where you're streaming DAPT with audio resources and you want to do client side mixing. Instead of including as base64, include as subsamples in the MP4 and have a way to reference them from the TTML. They could then be more efficiently encoded
Cyril: Another way is to have an audio track, and refer the TTML document to sections of the audio track
Nigel: Sounds plausible as an alternative
… Adaptation sets currently let you have audio main mix and audio with AD, but there isn't a way in DASH to have both and have the client side mix them
Future meetings
Nigel: Chairs and editors aren't available for the August 1 meeting
… Propose cancelling that one
Cyril: I may not make the July 18 meeting
<atsushi> in DST, meeting starts from midnight, but non DST it's from 1am.. ;)
Meeting close
Nigel: Thanks everyone, we're slightly over time, let's adjourn for today. [adjourns meeting]