Meeting minutes
This meeting
Nigel: Today we have some DAPT, some IMSC items.
… Any other business, or points to make sure we get to within those topics?
no other business
DAPT
Nigel: Regarding CR publication there has been some movement in getting the Security review completed.
Nigel: There has been some discussion, even up to 5 minutes ago
… It seems there is a conversation happening, which I hope we can conclude soon.
… For example there is a question about whether there is provision for data integrity so that
… it can be determined whether a file has been modified on the way to the processor.
… My response would be that is out of scope.
Cyril: I agree. You can do an MD5 or whatever.
Nigel: It should not be part of the file specification itself.
… He also answers the question about metadata manipulation.
Cyril: It's interesting, we have a good threat model now.
… It's nothing on top of TTML2.
… What do we need to do to call the security review done?
… Do we need them to tell us it's ok or something needs changing?
Atsushi: Usually yes for everything to be okay.
… This should be first of all not such a common review request for a data format.
… Mostly the discussion is focused on data handling and browsers.
… I suppose there might be some need for teaching them about our fundamentals.
… At this point I can't say what comment or background information we should provide.
Nigel: I sense that we should continue the thread until all questions have been answered,
… and then check back in with the reviewer to confirm they're happy for us to proceed.
… I am expecting some kind of yes/no from the reviewer though.
Cyril: I agree, continue the conversation and then ask them if we need to change the spec or if we are good to go,
… and then take on their response.
Nigel: I agree
… Any other points about the reviews or moving to CR?
Cyril: Not about that, but last time we talked about producing test vectors.
… We should maybe do a session and produce them.
Nigel: Yes
Cyril: Let's coordinate offline Nigel
Nigel: Yes let's do that.
… We have an Implementation Report but there's nothing to add - no change since last meeting.
Cyril: We said we wanted to add at-risk features.
Nigel: Yes we did, I haven't got around to doing it.
… I have an AOB I will raise about DAPT.
IMSC 1.3
Nigel: The only thing on the agenda is the Update namespace documents ticket
Update namespace documents w3c/imsc#589
github: w3c/
Nigel: I think we're waiting on Atsushi for how to update namespace documents
Atsushi: I'm still getting up to speed on this.
… I think I will open a PR onto the W3C repo directly and see what will happen
Pierre: Sounds like a good idea
Atsushi: I'm not sure what kind of DTD or supporting material we should attach to it.
… I suppose nobody in industry will expect material or DTDs from the namespace URLs, right?
… I had several identity-related libraries getting DTDs from namespace URLs
… but for our case in TT we just use them to define a namespace.
… I'm not familiar with the tools for implementations.
Pierre: I think that the requirement is that every time a namespace is created in a document
… that it be formally reserved, set aside, and the way to do that is by publishing a namespace document.
… Even though the chance of namespace reuse is very low, I understand this is a formal requirement of W3C.
Atsushi: If implementations will not look at the contents of the URL, we may
… not need to pay strict attention to its content.
Pierre: There's no DTD here, or XSL or anything like that. It's just a web page.
Atsushi: Then let me go and try and see what happens.
SUMMARY: @himorin to open a pull request to add the new namespace pages
Add support for subscripts and superscripts w3c/imsc#585
github: w3c/
Nigel: [summarises https://
… So the two main questions are:
… 1. Do we need to fix up TTML2 before referring to fontVariant?
… 2. How much do we care about what the UA actually does if it doesn't fully support this?
… Or if the font doesn't fully support superscript and subscript for every glyph in the super or subscript text.
Pierre: Overall I think this is a feature in flux in CSS, it changed even in the past couple of months.
… The semantic intent is pretty clear though.
… The fallback is very specific today, to HTML and CSS as currently implemented.
… My conclusion is not changed, that for the purpose of TTML and IMSC we can use fontVariant
… and things like imscJS will probably implement the fallback, but it's an implementation detail that
… might change in 2 months.
… Separately we can improve TTML2, but it's a separate issue.
Nigel: OK, second question: what is the range of acceptable behaviour for the UA?
Pierre: Looking at the current text...
CSS Fonts Level 4 Subscript and superscript forms
Pierre: [reads text from CSS spec]
… I think the intent is super clear
… The dispute in the browser world is not the semantics, but about whether or not the UA
… is expected to synthesise the font variant if it doesn't exist.
Nigel: OK, but there is no link through via TTML2 to this CSS spec text.
Pierre: You mean because the link is broken?
Nigel: Yes, but even if it weren't broken, it's informative and the TTML2
… text on tts:fontVariant has a lot less information in it.
TTML2 tts:fontVariant specification
Nigel: It doesn't hint about synthesis or fallback
Pierre: There's a semantic basis in the table
Nigel: That's what's broken!
Pierre: I can raise a PR to fix the semantic basis link to CSS.
… It links to CSS Fonts Level 3...
Nigel: It's missing from there
Pierre: No, it's in there.
Nigel: Did I make a mistake? Oh, it's just the section number that changed.
CSS Fonts Level 3 §6.5 Subscript and superscript forms
Nigel: Clearly they want to make some tweaks in Level 4 but this is fine.
… TTML2 doesn't even hint at anything except for use of the OpenType features,
… but CSS certainly does.
… I'd be a lot happier if either TTML2 or IMSC said something about fallback alternatives, either way.
… Basically, is it ok if a UA doesn't do anything when the sub or super variants are absent from the font?
Pierre: I think it's explicit on what it should look like, the fallback is an implementation detail I think.
Nigel: I could read it a different way, which is that the implementation should simply select the
… variant glyphs using the OpenType feature, and if they're absent, there's no behaviour specified.
… In other words, anything is ok.
Pierre: I think that's the point of contention in the browser community, last time I checked.
Nigel: OK, so what do we do?
Pierre: I think we should be happy with the semantic basis definition, and the syntax.
… I don't think anyone disputes what superscript and subscript mean
… I'd consider correcting the section reference in TTML2, but otherwise leave it as is.
… If CSS settles in a couple of months maybe we'll have a reason to revise it.
Nigel: I think that's a good starting point, and like you say there's time for us to keep watching and thinking.
Pierre: I will have to pay attention to it.
Nigel: [hunts for CSS issues]
… Too many to trawl through right now.
SUMMARY: Update the semantic basis reference in TTML2, keep watching for CSS updates on this feature
Nigel: It's w3c/ttml2#1277
Pierre: I'll prepare a pull request for that.
… Any comments on the IMSC pull request?
Nigel: No, the text is super simple
Pierre: If you can approve it then I'll fix up TTML2
Nigel: OK
Pierre: Super
AOB - Assembling timed transcript of e.g. DAPT
Nigel: Use case is, you have a TTML document that describes how new words get added over time.
… But then you want to present that in a non-TTML sort of view,
… As a complete document, where you do something like highlighting active text at a particular time
… during playback of related media.
… Like a transcript view.
… But then the source TTML might have content in different ISDs put into different `<p>` elements
… even if that text is all part of the same sentence, for example.
… Then when you assemble those `<p>` elements together the semantic connection between them
… doesn't exist and you get weird paragraph breaks in the middle of sentences.
… Really a problem for DAPT because of the timing structure we have defined.
… I wonder if, maybe not in v1, we need to add markup to allow transformation processors to
… understand that different block elements actually contain related text, e.g. part of the same sentence
… spoken by the same person.
… Has anyone else encountered this or developed any solutions?
Cyril: Like transcripts alongside videos that highlight the currently spoken text - like that?
Nigel: Yes, that sort of thing.
Cyril: What sort of relationship do you want to represent?
Nigel: Continuations really, where you don't want a line break.
Cyril: Could you use timing-based heuristics?
Nigel: Sometimes people leave gaps during sentences but you don't want a line break.
Cyril: It's subjective - I agree you would need to capture the author's intent.
Andreas: You are targeting a presentation out of scope of TTML, right?
… You are looking for something that would be a different presentation form.
Nigel: Strictly, yes.
Andreas: So you would want to add some metadata to give the flexibility to make it possible?
Nigel: I think so, yes.
Andreas: I can only think that metadata to relate TTML content together, unrelated to TTML layout,
… that you could use for whatever you want, could help. Some semantic unit ID or whatever, so you can make sense of it.
Nigel: Yes that's the sort of direction I was heading in.
… Thanks, that was a useful discussion.
Meeting close
Nigel: Thanks everyone, we're at time so let's finish.
… [adjourns meeting]