Meeting minutes
This meeting
Nigel: This may be the shortest ever meeting. There's only one agenda item, and I can't add to that right now.
… I've been in contact with Atsushi about the banner text on TTML2 1st Ed Rec, and he hasn't got an answer back
… from the team yet. Also Atsushi and Glenn have sent their regrets so we can't sensibly discuss that topic.
… Is there any other business?
Nigel: Checkpoint on IMSC: no feedback so far.
… 2 PRs to be merged soon.
Cyril: Quick question for the AOB: about the profile registry.
Nigel: Ok. Any other agenda items?
group: [no other agenda items]
AOB - Profile Registry
Cyril: Netflix is interested in linking from the profile registry to a web page where we describe the restrictions
… that we put on IMSC 1.1 Text Profile when we accept them for delivery.
… Is that acceptable for the profile registry?
… Initially my understanding was that it defines a 4cc for the codecs parameter to help clients determine
… if they can play or not before downloading documents.
… In this case we are talking more about an ingestion profile for delivery from studios.
Andreas: We also have the EBU-TT Part 1 profile listed there for contribution and exchange.
Cyril: That answers my question - it is not only for delivery to end user clients.
… That's very similar to what Netflix wants to register.
Nigel: That's not quite the same thing though - it is for published profiles
Cyril: There would be a profile designator and a URL, which would probably be public.
… What is the process to define it?
Nigel: Are you planning to specify an input constraint with a 4cc pointing to netflix and also im2t?
Cyril: No, it would be a set of constraints.
… So if I follow the steps in §3 and email the Chairman then it can be added?
Nigel: If you do that then I will do the next stage and put it to TTWG for agreement.
Cyril: Ok
Cyril: We would like vendors to document when they conform to Netflix requirements.
… We have recommendations and restrictions.
… Restrictions mean rejection of non-conformant documents.
… But an IMSC 1.1 Text profile document that meets restrictions but not recommendations would still be fine.
… For example there could be a special export saying "conform to netflix" who would add the identifier.
… They wouldn't have to say it's a Netflix compatible document.
… For example we don't allow animations in IMSC 1.1 documents.
… We can't ask people to update existing documents to include the Netflix profile identifier. It's a nice to have.
Andreas: I think in general it is super useful to have the profiles documented in the registry if they are in use in the wild.
… If people find the identifier they can resolve it and go back to what it means.
… It would of course be very helpful from the group's perspective if the documents are public available, though it is not a requirement.
… Because then if you find a document with this identifier then you can really guess what it means
… regardless of if you want to send it to Netflix or not.
Cyril: It's likely that we'll have core restrictions and then profiles for different languages but only one profile identifier.
… For example Japanese would have ruby but Arabic might restrict it.
… And then writingMode, direction etc for Arabic that would not be needed in Japanese, perhaps.
TTML2 2nd Edition CR Publication
Cyril: I'd like to understand how we will progress on the test suite.
… Is my understanding correct that the change section compared to 1st Edition, and there's a set of syntactical
… and a set of semantic only changes and we have something like 10 of each, roughly. Do we have to provide
… 1 test per change or are some not needed. Do we have any idea of the amount of tests we need to write?
Nigel: Yes we do. Glenn summarised this in last week's call so hopefully I scribed that.
… In summary, we have 2 categories of change - testable and untestable.
… The testable ones come partly after we introduced a working practice where all PRs would have associated tests.
… Those ones are okay. The others before that cut-off date need tests to be written.
Cyril: I see, from last week, 15 changes that need tests.
… Is there a master document? The IR is empty at the moment.
… For those 15 tests, or 11 that have already been created, if I look at the TTML2 tests repo and the pull requests since
… the publication of TTML2 that should give me all of them, right?
Nigel: Yes, I think so.
… Are you motivated to go and add them to the IR?
Cyril: Yes, I'm not sure how long that will last though (jokingly).
… Also we need to identify the gaps, right.
… I haven't done the exercise. I'm pretty sure some changes are in parts of TTML2 that we don't implement.
… That would be problematic if that is the case for everybody.
Nigel: True. Some of them might be for audio features, and BBC has an implementation that we may want to update.
Cyril: Did we update the XSD?
Nigel: I don't know.
… Good question.
Cyril: Some of the validation tools rely on the XSD at least flagging some things.
… For example adding 'ignore' value to the ttp:profileCombination attribute. That could be tested in the XSD.
Nigel: Yes, but a note of caution: not all the things that can be checked in XSDs are checked. Sometimes enumerated
… values are just listed as strings.
Cyril: I will try to put tests and names into the IR so people can start populating them.
Nigel: Thank you.
Cyril: The IR follows the same structure as the previous one. It is simple, we have to put a P or F.
… The IR only has validation results, whereas previously we had presentation results too. Is that intentional?
Nigel: I don't know, could be an accident. Without going through the changes I couldn't tell if they have presentation
… impacts.
Cyril: OK we need to go through that.
AOB - shear
Cyril: I am working with CSS WG people on shear, and some people have asked how it could be used in WebVTT.
… Gary, maybe we can sync on that?
Gary: Sounds good. Right now the only possible thing that exists is transform, and just allowing that is weird because
… we would only want to allow transform-skew. Only allowing a sub-property seems weird.
Cyril: Also depending on the semantics, you would have shear or lineShear. Is there a WebVTT property that applies
… to lines only?
Gary: No. I guess you would have a different class per line or have different lines over different cues.
Cyril: Yes, I'd have to think about it.
… I'm trying to revive this discussion, so if you're interested I'd appreciate some support in the threads when it starts.
Gary: Yes definitely.
Meeting close
Nigel: Thanks everyone, we've completed our agenda. Please let me and Gary know if you have any agenda items
… for next week before Tuesday if possible. [adjourns meeting]