Timed Text Working Group Teleconference

<scribe> scribe: nigel

This meeting

Nigel: We have one main issue to discuss for IMSC, though Mike sent his regrets.

Andreas: I won't be able to join next week.

Nigel: Okay lets get the views from here then and discuss it briefly without fully concluding
... until Mike has been able to contribute.
... That's all on IMSC. On TTML2 we have a number of issues to cover.
... I don't think there are any other main agenda topics, I can give an update under AOB
... on my trip to CSUN last week.
... There's been some progress on moving WebVTT to the next working draft but I don't
... think that's an agenda item for us today. You'll have seen David's encouragement by email
... to review and look at the dispositions.
... Any other business, or specific points to raise?

Dae: For IMSC 1.next we are introducing fillLineGap? Will this carry forward to TTML2?

Nigel: Yes we need the semantic in TTML2, and I believe that Glenn has begun some work
... on it in TTML2.

Dae: There's no support for it in CSS yet though?

Nigel: No, I did raise an issue on CSS.

Pierre: I started a wiki listing CSS features that are needed for subtitles. I encourage us to
... lobby for this.

Nigel: +1

Andreas: A question for Nigel as BBC representative: It's clear that BBC guidelines have a
... clear requirement documented for this feature. Could this be brought to the awareness
... of the CSS WG? I think it is nowhere more obvious than in the BBC guidelines. It is just
... an example but you made a strong point for it too.

Nigel: Yes absolutely, very happy to do that.

https://github.com/w3c/imsc/wiki/TTML2-features-for-inclusion-in-IMSC2

Pierre: That's the list of TTML2 features for inclusion in IMSC2.

Nigel: There's no pointer to which features are or are not doable in CSS?

Pierre: No, I started doing that with textEmphasis but I haven't with anything else.
... I can help with the mapping but ultimately Glenn is the expert there in terms of TTML2,
... so it requires his participation.

Glenn: If you're talking about CSS mapping I have no interest in doing that.

Dae: I don't mind volunteering to take a first go at this because most of the features have
... a straightforward mapping or fallback. I wonder how robust the fallback has to be?

Glenn: My solution to mapping to HTML is to map to SVG in HTML.

Nigel: To an extent that does not fully resolve the problem since CSS affects the presentation
... of text in SVG in any case and may be used to make it more accessible.

Glenn: I don't recognise that, metadata can be added to SVG to make it accessible.

Nigel: Since IMSC2 is a subset of TTML2 then any features that we want to be in IMSC2 need
... to be in TTML2. If we need IMSC2 to be mapped to CSS then we need to make sure the
... fallbacks and limitations are clear.

Glenn: We decided some time ago to make HTML/CSS mapping orthogonal to TTML2
... development.

Dae: I will aim to do a first draft of the CSS mapping before next week's meeting.

Glenn: I don't disagree that it is useful to know where the challenges are. If there are tweaks
... that can be made to make things a little easier I don't generally object to doing that.

Nigel: We need to pragmatic here - if we are forced to wait for a normative reference to
... a CSS spec for everything then we will be waiting for a long time.

Dae: I think this is about a fallback in some cases.

Pierre: For example what is the CSS equivalent of fontShear?

Glenn: There is none.

Pierre: We should note it then.

Glenn: We have not done that in the past. For example in TTML1 XSL-FO does not support
... all the semantics, and there's a note that declares that such unmappable semantics exist.
... We did not let that gate us for TTML1.

Pierre: In the scope of IMSC1 features there is in fact a clear mapping via XSL-FO to CSS.
... XSL says what the differences are between it and CSS. That's the case for TTML1 now and
... it made the mapping clear and easy to do.

Glenn: It's under §9.3.3 under synchronic flow processing, a note about there being no
... XSL counterpart.
... We've had to go beyond XSL-FO and CSS in TTML2 due to new requirements that
... still are not satisfied in CSS land.

Andreas: What we need to acknowledge is that since TTML1 started as a standard the
... importance of CSS mapping has increased in general in the domain and for members and
... for W3C. For IMSC1 and IMSC2 there may be the situation where the industry will not
... support a feature without a clear CSS mapping, so I can support that for everything we
... want to include especially in IMSC we should do everything we should do to make it
... work with web technology.

Pierre: Are you opposing doing the mapping work Glenn?

Glenn: I am not opposing it, I just don't think it is a necessary part of TTML2 and can be
... done elsewhere.

Pierre: So you will not object to that or to PRs on TTML2?

Glenn: Sure, I will consider PRs on TTML2.
... The other question that is deeper is that during that mapping process there may be a
... suggestion to make changes in TTML2, and that's reasonable, but we have to ask ourselves
... the practical question if waiting for that to be completed could become a blocking or gating factor for wide review.
... If during the WR process progress is made on this then we can look at that and decide
... if we want to make any changes as we come to CR to accommodate that.

Nigel: [draws this discussion to a close]

IMSC

https://github.com/w3c/imsc/issues/213

Nigel: I am not sure what the outstanding points are with this that need to be discussed.

Andreas: I think the last edit from Pierre would be fine. Nigel you brought in an extra
... issue which is how far you give a recommendation or requirement for how far foreign
... namespace content should be defined. My view is this is not in scope of IMSC or TTML
... so would not be necessary, but it is just a recommendation so I wouldn't object to it.

Nigel: It's even weaker than that - it's an informative note only to say that it is a good idea
... to define a content specification but I agree that we cannot force people to make good
... implementations!
... It's in a pull request:

https://github.com/w3c/imsc/issues/217

Pierre: Now Mike has told me he wants to be specific about exactly which elements
... support foreign namespace elements and which do not, but I do not want to rewrite
... TTML1!

Nigel: There are 3 new things since last time:
... 1. If we should remove the exception for the tt element?
... 2. Clarify that foreign namespace attributes are permitted on all TTML1 elements.
... 3. Add an informative note about good practice for content specification.

Pierre: I think there's a dispute that has not been resolved where Mike wants to introduce
... consideration of XML Schema validation.

Nigel: I think that's further than we should go.

Andreas: First I thought we do not need to be as specific as in TTML1 but now when I look
... at it maybe it would be good to document it wherever it is, because for example
... <style> does not allow <tt:metadata> as a child element, which differs from what the
... EBU group does, and has done for years, so there are people making mistakes on this.

Pierre: So you're supporting Mike's idea of explicitly listing the elements where metadata
... is permitted?

Andreas: I can understand not wanting to put too much in the IMSC spec but somewhere
... it should be documented.

Pierre: My reluctance is simply that it requires interpretation of TTML1, which I typically
... recommend we don't do, but if we feel it is useful and Mike thinks so then maybe it
... is enough.

Nigel: I really like the way it is done now by referring to Table 3 in TTML1.

Pierre: It is wrong though because for example <head> permits metadata. Maybe we should
... for simplicity just list the elements directly.

Nigel: OK, I'm fine with listing in an informative note the elements. That's probably a service
... to the world.

Pierre: In that case I'm happy to do it.

Nigel: The second point is to be clear that foreign namespace attributes are permitted on
... all TTML1 elements.
... I think if we are going to summarise what TTML1 permits we should do it accurately.
... I went to the trouble of checking TTML1 and in fact every element defined in that spec
... permits foreign namespace attributes.

Pierre: The risk is that people should go and read TTML1.

Glenn: I agree that we should not paraphrase TTML1, it's okay to have a generic note saying
... to see TTML1.

Nigel: Summarising, there's a concern that we should not list things that are defined in TTML1.
... For me, I think we should either summarise correctly or not at all.

Pierre: I think you and Mike are agreed on that.
... I will list all the elements that permit foreign namespace metadata and change the lsit
... of elements in TTML1 that permit foreign namespace attributes.

Nigel: Moving on to the note about content specifications, the note I have added is:
... "For validation purposes it is good practice to define and use a content specification for all foreign namespace elements and attributes used within a Document Instance. "
... The reason I think we should do this is from going back to the text in TTML1 about
... validation, you may end up with non-pruned foreign namespace elements that should
... be validated against a content specification, and this provides a small hint to plug that gap.

Andreas: I will not object, but I would prefer not to do this. It could be good practice also
... to ignore unexpected content.

Pierre: I will add that paragraph, update the PR. If Mike is happy then we go ahead. If not then maybe the solution is to remove that paragraph.
... if that also satisfies Mike.

Nigel: Ok I wouldn't object to that if it gets us to consensus.
... OK that's all for IMSC, let's take 2 minutes and come back for the next hour on TTML.

TTML

Dae: I'd like to clarify if we are addressing horizontal review comments before CR?

Glenn: When I made that comment I did not plan to have to resolve all HR comments prior
... to going to WR. I did not mean I would not address them if possible. I think it is still an
... open issue about if we need to resolve them prior to going to WR.

Nigel: We will need to make a call when we want to resolve to publish the WR for WD.

Pierre: I recommend assigning issues.

Glenn: I don't object to assigning issues to get more people working on them however I
... will need to do the merge so that will depend on the effort I have to do the editorial tasks.
... I have no objection to others preparing PRs. I might delegate the merge task to others if
... time is limited. I need to have oversight on the final text.

Nigel: I don't think we should have a blanket policy on pushing all HR issues out to after WR.
... We need to be able to take it on a case by case basis.

Pierre: For issue 233 on HDR that's a blocking issue for Movielabs.
... I'm happy to prepare a PR for it.

Glenn: Sure.

https://github.com/w3c/ttml2/issues/233

Pierre: I will prepare a PR for that.

Nigel: I expect a BBC proposal also, for the non-absolute luminance case.
... I've assigned that issue to you and me, Pierre.

Dae: If anyone else feels there is a specific issue that we must resolve as soon as possible
... then it would be good to make that known.

Pierre: Assigning issues is an approach that works well, where the assignee needs to propose
... a disposition.

Nigel: By the way I have one, the audio description one:

https://github.com/w3c/ttml2/issues/195

Pierre: We need to document any deferred issues and agree them in the group based on the rationale.

Nigel: On the AD issue I have done the pre-editing work to define the semantics and Glenn
... is in progress on the editing.

Pierre: For the TTML1 issues that flow into TTML2 also then in my view they all must be
... resolved. They're bugs.

Glenn: I have probably different criteria for bugs.

Pierre: There are a bunch of "discussed and agreed" issues.

Glenn: They're well up on my priority list and I'm trying to get through them whether they
... are labelled as bugs or not.
... I'm working on the 58 issues marked TTML2WR but I have no objection to anyone
... else assigning themselves to an issue.

Pierre: An efficient solution is not to ignore issues and respond to them sooner otherwise
... things will get worse in the future and people might get angry.

Glenn: I'm not trying to make any decisions for the group by the way, I'm just labelling
... what I am working on.

Pierre: I have to leave now.

Nigel: By the way Andreas left before the TTML agenda item too.

Dae: I would like to raise the bopomofo issue

https://github.com/w3c/ttml2/issues/253

Glenn: This refers to a particular script which is used in Taiwan to annotate Chinese
... in pedagogical works. I would not expect it to be seen in subtitling except possibly in
... educational programmes for children. It is only really used when teaching children in
... Taiwan how to read Chinese.

Nigel: I think we should put this on hold until Richard's request to @bobbytung gets a
... response, and act on that response when it comes. It may be that we decide this is not
... high enough on our priority list and we choose to go ahead without Bopomofo support.
... I did want also to discuss the em units issue:

https://github.com/w3c/ttml2/issues/216

Glenn: On that particular issue there is an editorial point on the text which I would like to
... be opened as a new issue.
... That one wasn't an early merge.

Nigel: That's strange, I commented that it was an early merge and marked as one, but the
... history doesn't show that. I'm a bit confused about that.
... It turns out that Pierre and I had similar editorial comments, so I guess we can raise those
... as new issues.
... There may be some consensus on the 3D and images issue:

https://github.com/w3c/ttml2/issues/239

Nigel: The last points by Mike, Pierre and I are that we do not need to have left/right images.

Glenn: On the point about using condition I would like to add named parameters there.

Nigel: I don't think that's necessary but I would not object to it.

Glenn: I can do that very quickly and post a PR that does that and see what comments there are.

Nigel: The only reason against doing that is if it generates additional test cases that we do
... not actually need. Otherwise I don't have a problem with that.

Glenn: I'll add a comment to the effect that this is what I will do.

nigel: The next issue I listed was the TTML1 pull request:

https://github.com/w3c/ttml1/pull/233

Nigel: This is one of Andreas's.

Glenn: We had the whole issue of whether padding should be on the inside or the outside of regions.
... I see this is a note. I need to do more than just add this, I would need to add an errata
... info and change info, but I see you made a suggestion.
... So, mutatis mutandis, we can mark this as discussed and agreed?

Nigel: I think so.

Glenn: That means "the necessary changes having been made"

Nigel: I've logged that on the issue.
... What about the open pull request?

https://github.com/w3c/ttml2/pull/273

Dae: Richard raised issues 274 and 275 based on that.

Nigel: Oh right I did not realise the relationship was there, there's no tracing in the issues or the pull request.

Dae: I think there's another one aside from 274, it might not be 275.

AOB

Nigel: Last week I went to the CSUN annual conference on assistive technology and presented
... on AD in TTML2 (current status anyway, and desired end goal) and also on live subtitles
... in TTML via the EBU-TT Live work.
... There was a really good and engaged audience for the AD in TTML2 talk, and in fact
... there were a number of AD talks over the conference. There turns out to be a use case
... for pan, and also for SSML for text to speech, for example to disambiguate different ways
... of processing a number like 1876 that could be a year or a number.

Glenn: I do not want to embed the whole of another language into TTML2!
... Also it would be difficult where they have created new elements and put attributes on those.
... What I have done is to create new attributes only to keep it consistent with how TTML does things.
... We have very strongly avoided elaborating our element vocabulary.

Nigel: Understood. Anyhow I thought it is worth mentioning this may be a good time to take advantage of the groundswell of good feeling.
... Thanks everyone. [adjourns meeting]