Timed Text Working Group Teleconference

trackbot, start meeting

<trackbot> Meeting: Timed Text Working Group Teleconference

<trackbot> Date: 01 December 2016

<scribe> scribe: nigel

This meeting

nigel: For today, I'd like to cover the TTML Profiles registry, maybe the Unicode ticket, and the Security & Privacy section for TTML2.
... AOB?

mike: I'd like a minute on the IANA registry and the page that supports that, and also
... the post from Pierre about imscjs.

nigel: Sure, we did actually discuss imscjs a little last week though it's not well minuted.

TTML Profiles Registry

nigel: This is about the issues and PRs etc, and if we should update it now.

mike: Is the IANA registration complete?

nigel: Yes it was updated on 2016-09-30

-> http://www.iana.org/assignments/media-types/application/ttml+xml

nigel: It refers to the profile registry document.
... Now looking at the open issues/PRs:

-> https://github.com/w3c/tt-profile-registry/pulls

nigel: This is the moment to say that the open issues/pull requests are all the work to be done.

mike: I think the passage of time is sufficient here.

nigel: I concur.

group: [no objection]

mike: OK I'll go ahead and push this up towards the top of my list.

nigel: Thank you!

IMSC (plus Unicode ticket)

action-478?

<trackbot> action-478 -- Nigel Megitt to Draft a liaison to hbbtv requesting further information and proposing an option e.g. to extend imsc 1 to allow signalling of background height on span, and request timelines etc. -- due 2016-09-26 -- PENDINGREVIEW

<trackbot> http://www.w3.org/AudioVideo/TT/tracker/actions/478

nigel: This is now sent and I updated the action with the archive link to the liaison that was sent.

close action-478

<trackbot> Closed action-478.

action-487?

<trackbot> action-487 -- Nigel Megitt to Send generic liaison to atsc, arib, mpeg, dece, cta, dash-if, dtg, hbbtv, smpte. -- due 2016-11-24 -- PENDINGREVIEW

<trackbot> http://www.w3.org/AudioVideo/TT/tracker/actions/487

nigel: Same applies here - the liaisons were sent, thanks to Pierre and all for helping edit
... the text. Links to the liaisons as archived are in the action.

close action-487

<trackbot> Closed action-487.

nigel: Apologies to Mike for the confusion about the SMPTE one, which was my mistake.

action-488?

<trackbot> action-488 -- Nigel Megitt to Send liaison to dvb replacing first sentence from generic response with acknowledgement sentence. -- due 2016-11-24 -- PENDINGREVIEW

<trackbot> http://www.w3.org/AudioVideo/TT/tracker/actions/488

nigel: I did this, and logged the message archive link in the action.

close action-488

<trackbot> Closed action-488.

nigel: One thing to note is I bumped the response date from 1 Jan to 8 Jan to accommodate my lateness in sending those messages. Apologies if that causes any difficulties!

action-485?

<trackbot> action-485 -- Thierry Michel to Ask richard ishida for assistance in unblocking the unicode ticket 8915. -- due 2016-11-10 -- PENDINGREVIEW

<trackbot> http://www.w3.org/AudioVideo/TT/tracker/actions/485

nigel: There has been an update on the Unicode ticket

-> http://unicode.org/cldr/trac/ticket/8915#comment:8

pierre: I think we have the scripts to produce one of the things they are asking for, the delta.
... Maybe we should go through the questions now?

nigel: I think that would be useful, quickly.

pierre: Question 1...

nigel: Does anyone here know the difference between supplemental and exemplar sets in Unicode?

pierre: I'm not sure, we maybe need to dig deeper or ask for clarification. This may not be a question for us to answer.

nigel: Agreed.
... It's good they're considering _how_ to do it rather than _if_.

pierre: Exactly, that is encouraging.
... Question 2: I think the answer is "yes" to both questions.
... I think the Unicode tables try to be statements of fact rather than mandatory. And that's all we need, a description of the characters that are used for captions in a particular language.

glenn: [speculates that supplemental sets are additional characters but examplar sets are lists of existing characters for a particular context]
... So I think the answer for question 2 is yes to a and yes to b.

nigel: +1

pierre: Question 3: We definitely have created that delta set and can point to it.

<scribe> ACTION: pal Update the Unicode ticket 8915 with a link to the deltas between what we recommend and what is already in main and aux sets. [recorded in http://www.w3.org/2016/12/01-tt-minutes.html#action01]

<trackbot> Created ACTION-489 - Update the unicode ticket 8915 with a link to the deltas between what we recommend and what is already in main and aux sets. [on Pierre-Anthony Lemieux - due 2016-12-08].

pierre: Maybe they are saying that if characters are in use for captioning in a language then
... it should be added to aux for that language. That would be fine by me.

nigel: Q 3a may be that they have found a mistake?

pierre: There are two parts to 3a. We can provide the background behind the selection, which
... is indeed an empirical examination of real data. The other part, about Hebrew and other
... latin scripts, I will follow up with someone who knows about this.

glenn: I agree with the question, it does look extremely odd. Also it may cause trouble having characters with different directions.

pierre: The basic latin set is included all the time even with Hebrew, so the question is why
... is a special case made for U+0178 : LATIN CAPITAL LETTER Y WITH DIAERESIS?

nigel: Looking at the list, Hebrew includes Latin Extended-A which has other characters
... as well as U+0178.

pierre: It sounds like either there is a mistake - I will check - but also explain that all the
... Latin Extended A characters are included (plus other Latin characters) in all the locales.

glenn: I don't believe it includes all Latin characters defined in Unicode.

pierre: I should have said the "base set" (whatever it's correct name is).

glenn: Yes, Vietnamese latin characters are not included for example.

pierre: Should I pass draft text to the group before updating the ticket?

nigel: I think you can go ahead and update the ticket directly.

pierre: Thanks for keeping tabs on the ticket by the way.

nigel: No worries. Thanks to Richard for prompting them - that really helped.

action-485?

<trackbot> action-485 -- Thierry Michel to Ask richard ishida for assistance in unblocking the unicode ticket 8915. -- due 2016-11-10 -- PENDINGREVIEW

<trackbot> http://www.w3.org/AudioVideo/TT/tracker/actions/485

nigel: Since this is unblocked, I'll close this.

action-485: [Meeting 2016-12-01] This seems to be unblocked - thank you Richard! Closing.

<trackbot> Notes added to action-485 Ask richard ishida for assistance in unblocking the unicode ticket 8915..

close action-485

<trackbot> Closed action-485.

nigel: Do you want a minute on imscjs?

glenn: I would like to hear more about it.

pierre: This is a javascript library intended for rendering IMSC 1 documents into HTML5
... as faithfully as possible to the spec.

-> https://github.com/sandflow/imscjs

pierre: There are three main parts, parsing, turning into ISDs and then mapping the ISDs into HTML5.
... That last part could be replaced by other renderings.
... There are also links to test sample renderings from TTML1 and IRT tests.
... This work is sponsored by MovieLabs.
... This is an initial beta release, intended to be substantially complete in API and in
... functionality, and the goal is to walk towards a 1.0 release in the next couple of months.

andreas: I also tested it and found it helpful. I did not find any error yet, and everyone I
... told about it said it is really helping for TTML implementation. Another thing that could
... be relevant is that very easily you can produce images for all of the TTML samples that
... are in the test suite, so therefore it could be helpful for this group, and you could think
... about providing images together with the test suite.

pierre: It works only in Chrome and Firefox because of security issues, but from Firefox
... you can actually export the sequence of images associated with each ISD.

andreas: It is also helpful for TTML to HTML mapping, so we also benefit from that.

glenn: I agree with Andreas, it's very nice to have this available.
... Did you create any new test content or did you use the former TTML test suite?
... Also have you had a chance to document any rough spots in terms of trouble mapping
... to HTML or CSS that required giving up formatting semantics or other things like that?

pierre: I've been using for test the TTML1 test vectors that are compatible with IMSC1,
... especially for timing semantics that has been extremely helpful. Also the IRT test vectors
... because they come with nice expected renders, so that's been nice for finding bugs.
... Ultimately I would like to come up with really good IMSC 1 specific test vectors. I have a
... question for W3C about copyright etc and how they would feel about turning TTML1 tests
... into IMSC 1 tests.
... On the second question so far I think in the case of IMSC1 there's nothing that cannot
... be mapped to HTML, but time will tell! I've seen some interesting things with unicode bidi
... and embed, and need to see if there's a spec or a browser issue - that will take more time.
... One of the IRT test vectors has a discrepancy.
... Another one is xml:space="preserve" which is also something where we might find surprises.
... Overall, the two features that required the most work were linePadding and multiRowAlign
... because both of those features require determining line breaks when there are no specific
... line breaks.

nigel: What about box-decoration-break: clone for linePadding?

pierre: The issue here is multiple nested spans with different background colours. The clone
... works great if there is only a single background colour for the entire background of the
... p or the span.

glenn: On multiRowAlign did you end up using display: inline-block ?

pierre: Yes I think so, after finding line breaks and inserting br elements where needed.

glenn: I'll have to take a look at it with some examples - these are tricky areas for sure.

pierre: The strategy is to do a first pass putting every character in its own span, then check
... the vertical position of each, and when it changes then insert a hard break.

mike: Comment: I had the opportunity to throw some DECE film content at it that was
... coded to DECE specs, and it looked like it was doing the right thing, without looking
... in extreme detail. That was encouraging, with 300-400kb of XML.
... Question: There's a partial implementation of IMSC-1/EBU-TT-D in dash.js and I wondered
... if you had some thoughts about integrating into that?

pierre: Yes, in fact I have been touch with one of the main developers of dash.js TTML code,
... and his comments were 1. How are images supported? and 2. Why are IE and Edge not supported?
... In fact images are supported, using an image resolver.
... I have not had time to test with IE and Edge, and I was pleasantly surprised how
... far IE11 has come, and there's a pull request to make 2 changes that will make the
... library compatible with both. The goal was always to make it compatible with Edge.
... I expect to merge that very soon, and then update the README. It should be compatible
... with the latest version of Edge, Safari. Chrome, Firefox and IE11.
... It turns out there is only one CSS feature not supported by Edge and it was really
... easy to get around it.

TTML2

nigel: I just want to mention Security and Privacy. We currently don't have a section on this,
... and I've discussed with David Singer and Glenn and this has led to issue 224 being raised.

-> https://github.com/w3c/ttml2/issues/224

nigel: We have three examples now, but any others would be welcome - please contribute to the issue.

glenn: I thought of another, the use of xlink to cause browsing away to other content.

nigel: I've added that to the issue.
... We're out of time for today (slightly over) and we've covered the main agenda topics. Thanks everyone! [adjourns meeting]