{minutes} TTWG Meeting 2018-05-17

Thanks all for attending today's TTWG meeting. Minutes can be found in HTML format at: https://www.w3.org/2018/05/17-tt-minutes.html

Reminder: we have no call on 24th May

In text format:


   [1]W3C

      [1] http://www.w3.org/

                Timed Text Working Group Teleconference

17 May 2018

   See also: [2]IRC log

      [2] https://www.w3.org/2018/05/17-tt-irc

Attendees

   Present
          Nigel, Cyril, Glenn

   Regrets
          Thierry

   Chair
          Nigel

   Scribe
          nigel

Contents

     * [3]Topics
         1. [4]This Meeting
         2. [5]TPAC 2018
         3. [6]TTML2 items labelled as Agenda
         4. [7]Rename and refactor module subsections. ttml2#591
         5. [8]Generic processor conformance too strict. ttml2#673
         6. [9]Move processor conformance out of #validation
            ttml2#689
         7. [10]The #extent-measure designation includes the value
            auto. ttml2#690
         8. [11]Move processor conformance out of #validation
            ttml2#689
         9. [12]Use of PNG image format. ttml2#694
        10. [13]Use of HDR images. ttml2#695
        11. [14]Negative feature designators aren't future proof.
            ttml2#697
        12. [15]Rewrite version 1 features as references to
            [TTML1]. ttml2#763
        13. [16]Disallow `<length> left` and `<length> right`
            syntax. ttml2#711
        14. [17]Definition of implicit inline region ttml2#727
        15. [18]Featurize offset times w/o metric. ttml2#752
        16. [19]Meeting close
     * [20]Summary of Action Items
     * [21]Summary of Resolutions
     __________________________________________________________

   <scribe> scribe: nigel

This Meeting

   Nigel: Today, I don't think we have anything to discuss for
   Charter, I do need to quickly
   ... cover TPAC, then I expect the main part of the meeting to
   be taken up by the TTML2
   ... issues marked as agenda.
   ... Any other business, or particular topics to cover other
   than those TTML2 agenda issues?

   Cyril: No

   Glenn: No

   Pierre: [silence - getting coffee?]

   Andreas: No

   Nigel: Great, let's get going.

TPAC 2018

   Nigel: The draft schedule has been published, as linked on the
   agenda:

   [22]Draft TPAC 2018 schedule

     [22] https://www.w3.org/2018/10/TPAC/schedule.html

   Nigel: This has us meeting on Monday and Tuesday, and also has
   the CSS WG and Media & Entertainment IG
   ... meeting on Monday (CSS WG also on Tuesday).
   ... The AD CG has time on the Thursday also.
   ... My question to you guys is do we want to request a change
   to the schedule?

   Andreas: I definitely think we should request a change in the
   agenda. The M&E especially
   ... has important stakeholders of TTML and so on, and a really
   good opportunity to discuss
   ... TTML specific issues, and as a representative of a
   broadcast organisation I plan to join
   ... the complete meeting of the M&E IG.

   Pierre: The fact that it is at the same time as CSS WG is a
   good thing because it guarantees
   ... that CSS folk will be there so we can ask for an hour joint
   meeting and have reduced
   ... likelihood that we will not be able to get them to join.
   ... M&E IG is a different matter.

   Nigel: I've also been talking to my colleague Chris Needham who
   is one of the chairs of
   ... the M&E IG about them moving potentially.

   Andreas: My impression of CSS WG is that they have more than
   enough to discuss in their
   ... two days, so it may be easier to schedule a joint meeting
   if we do it on another day.
   ... The attendance we had last time in the joint session was
   sufficient to bring in the major points.

   Nigel: I think what I'm hearing is we want to avoid a clash
   with M&E and to arrange a joint
   ... meeting with CSS WG however that is most conveniently
   achieved.
   ... The next question is about our flexibility.
   ... Could we for example ask to meet on Tuesday and Friday?

   Pierre: I could not be present for an entire week at TPAC.

   Nigel: Implying Mon/Tue or Thu/Fri?

   Pierre: Yes, with a preference for Mon/Tue

   Nigel: Any other constraints on us moving the meeting dates?

   Andreas: No, I should be available on Mon,Tue, Thu and Fri.

   Nigel: Okay if there are no other views that's enough for me to
   take back.

TTML2 items labelled as Agenda

   Nigel: We have 15 open items labelled as agenda in ttml2:

   [23]TTML2 agenda items

     [23] https://github.com/w3c/ttml2/labels/agenda

Rename and refactor module subsections. ttml2#591

   github: [24]https://github.com/w3c/ttml2/issues/591

     [24] https://github.com/w3c/ttml2/issues/591

   Glenn: I don't think we need to do this.

   Nigel: I see that it's about moving sections around so that
   common text does not need
   ... to be duplicated.

   Glenn: This would cause specref links to lose contextual
   information.

   Nigel: The second part of the issue is that §10.2.1 should
   apply to the audio attribute
   ... vocabulary whereas it currently looks like it does not.

   Glenn: This will make the ToC deeper and cause a wide scale
   renumbering which would
   ... invalidate existing references from other specs like IMSC
   1.1. I've been trying to maintain
   ... section numbering referential integrity. Of course we've
   had to add new numbers in the
   ... 10.2 section so those did get renumbered somewhat. I can
   see the logic of putting those
   ... together. It seems like the only reason to do this is to
   allow 10.2.1 to apply to both.

   Nigel: The audio attribute section does not refer to 10.2.1 and
   it is scoped out by the
   ... section structure. This is the problem.

   Glenn: [thinks] What about §10.2.1 is tts-specific?

   Nigel: 10.2.x includes the style attribute and everything in
   tts namespace,
   ... whereas 10.3 includes tta namespace.

   Glenn: This is editorial only.

   Nigel: Are you sure there's nothing substantive implied by the
   sectioning?

   Glenn: Yes, there's nothing substantive.
   ... I could put a note in the prologue for section 10 at the
   end describing why we have a
   ... separate section for audio style properties and that these
   generic style features apply
   ... there as well. I don't think there's a problem to solve
   here.

   Nigel: Why do we have 10.3 at all? Why not just run them into
   §10.2?

   Glenn: We could do that, but they would all apply at the top of
   the list before the tts:
   ... attributes, using alphabetical ordering.

   Cyril: Since this is editorial can we resolve this in CR2 if we
   need to?

   Nigel: If we can solve it now quickly that would be good. We're
   doing this now.
   ... Any other ideas?

   Glenn: Moving the tta: attributes into 10.2 would work

   Nigel: Works for me.

   Pierre: I don't object.

   RESOLUTION: Move the tta: namespace style attributes into §10.2

   github-bot, end topic

Generic processor conformance too strict. ttml2#673

   github: [25]https://github.com/w3c/ttml2/issues/673

     [25] https://github.com/w3c/ttml2/issues/673

   Nigel: The summary of this is that it is reasonable to have a
   content profile prohibit use
   ... of features that are mandatory for generic processor
   conformance.

   Glenn: I don't think we can make them optional because they're
   mandatory in TTML1.
   ... We have another issue that's related, about making
   profile-version-2 optional: #683
   ... In pull #755 I proposed some language in the claims section
   that allows for conformance
   ... without supporting the mandated features. I think that
   applies in this context too.

   Nigel: Yes, I see that.

   Glenn: You're saying you might have a profile that doesn't
   support things that seem to be
   ... mandated by the official minimum profiles, which is not
   unreasonable and has been done,
   ... for example EBU in excluding the profile features. So your
   additional comment will be
   ... dealt with by pull #755.
   ... Basically a processor can be a ttml processor but cannot
   claim conformance as defined
   ... by the conformance clauses.
   ... It can claim generic conformance but not one of the two
   special ones.

   Nigel: But §3.2.1 bullet 4 requires support for everything
   listed as M in §E.2.

   Glenn: That doesn't mandate support for e.g. #profile.

   Nigel: The way I read it is does, because step 4 must be
   satisfied, and doesn't refer to
   ... any profiles, just the specification, and it clearly
   relates to things listed as M in §E.2, which
   ... includes `#profile`.

   Glenn: I see what you mean. It's possible that the note is
   wrong in the context of such a
   ... content profile not requiring support for everything listed
   as mandatory.

   Nigel: Good, we're on the same page in terms of the issue at
   this point.

   Glenn: Either you have to implement everything even if use is
   prohibited by a profile, or
   ... we somehow relax that language to permit the creation of an
   implementation that doesn't
   ... require support for all those mandatory features.

   Nigel: Yes

   Glenn: An option is to have a lower case ttml processor
   conformance state that does not
   ... include Generic Processor Conformance.
   ... Back in TTML1 days we had a huge discussion about minimally
   required features. The
   ... first chink in that was when EBU-TT was created. Just a
   historical fact.
   ... Now with IMSC we seem to be following the same path.
   ... It looks like we'll need some more thinking on this.

   SUMMARY: Issue discussed and mutual understanding improved,
   more thought needed.

Move processor conformance out of #validation ttml2#689

   github: [26]https://github.com/w3c/ttml2/issues/689

     [26] https://github.com/w3c/ttml2/issues/689

   Glenn: This is for Pierre who has asked additional questions on
   this.
   ... [summarises comments on the thread] Pierre, do you still
   think there's a problem?

   Nigel: We can't hear from @palemieux right now, will have to
   come back to this one.

The #extent-measure designation includes the value auto. ttml2#690

   github: [27]https://github.com/w3c/ttml2/issues/690

     [27] https://github.com/w3c/ttml2/issues/690

   Glenn: Looks like Pierre responded...

   Pierre: Yes, the note would really help. This was caused by the
   fact that the terms are
   ... identical but mean something different depending on the
   context.

   Glenn: OK

   RESOLUTION: Add a note as per the comments in the issue.

Move processor conformance out of #validation ttml2#689

   github: [28]https://github.com/w3c/ttml2/issues/689

     [28] https://github.com/w3c/ttml2/issues/689

   Pierre: I think there is still a problem. We have a conformance
   section, §3, where conformance
   ... of various kinds of processors is defined, and suddenly in
   the validation feature there is
   ... a definition of a validating processor and I don't know why
   this is done but not in §3.

   Glenn: There is no such thing as a validation processor, though
   I did discover two non-normative
   ... uses of that in the Introduction, which I can remove. A
   validating processor is a sub-processing
   ... step of a content processor, either a transformation or a
   presentation processor. There
   ... is no such thing as a standalone validation processor. It's
   an optional bit that gets turned on.
   ... So there's no need to add a separate definition.

   Pierre: An implementation today can be either a transformation
   or a presentation processor.
   ... Why not define a validation processor and an implementation
   could be all three at once?

   Glenn: There's no need to.
   ... The only reason for doing so is for the purpose of making
   claims. Nobody is going to make
   ... a validating content processor that is only a validation
   processor in my opinion, and if
   ... they wanted to do so it would be a validating
   transformation processor with a yes/no outcome.
   ... There is no need to add a new class.

   Pierre: Then lets remove the words "is a validating content
   processor".

   Glenn: But that's the point of that section...

   Nigel: I'm struggling to see the problem here.

   Pierre: The `#validation` feature designators is unique amongst
   all the feature designators
   ... because it imposes more than the processing of vocabulary
   but also conformance to
   ... a particular validating content processor definition. My
   suggestion is that if there is really
   ... such a thing as a validating content processor then it
   should be defined in the section
   ... on conformance. Either way "is a validating content
   processor" should be removed, and
   ... if there is a validating content processor then it should
   be added to §3.

   Nigel: Okay point 1, why should we not remove bullet 1 from
   `#validation`, i.e. remove
   ... the words "is a validating content processor".

   Glenn: This is not the first feature designator that does not
   define a syntactic construct,
   ... e.g. lineBreak-uax14.
   ... The current specification of ttp:validation does not
   mandate support for a validating content processor,
   ... it defines semantics that apply to a validating content
   processor but does not require
   ... it to be one. I could just move support requirement to
   ttp:validation.

   Nigel: Isn't support for `#validation` something that implies
   that the supporting processor
   ... is a validating processor?

   Glenn: It would mean that logically there is no effect.
   ... The `#speech` designator uses the same language. We don't
   define a conformance
   ... section for a speech synthesis processor.

   Nigel: They aren't identical, you could use the same approach
   for speech to validation.

   Glenn: There are two things I could do:
   ... I could use more of the approach taken within `#speech`.
   ... Or I could use the phrasing from speech to make it explicit
   that validation is an optional
   ... component of any content processor.

   Pierre: I still don't understand - the current definition says
   "if it is a validating content processor"
   ... I follow the link to the definition and it doesn't tell me
   anything.

   Glenn: That's a good point, which I was thinking about. The
   real intent is to implement the
   ... semantics of §5.3.1.

   Pierre: Why not just say that, instead of "is a validating
   content processor" just say
   ... "implements the semantics of §5.3 Validation"?

   Glenn: I could do that, I would prefer to do it in the
   definition of the validating content processor,
   ... because then I could do the same for the speech processor
   to make them congruent.

   Pierre: Then the definitions section starts to look like a
   conformance section, which is my
   ... point at the end of the day.

   Glenn: You could say that for everything. We are not going to
   create a conformance section
   ... for each of the defined features?

   Nigel: Would you accept a change to the definition Pierre?

   Pierre: No the feature should just refer to §5.3 directly.

   Glenn: [scribe paraphrases: this is about consistency across
   the spec]

   Nigel: The question is Pierre would you be able to accept the
   reference intermediated through the definitions section?

   Pierre: I would need to study it further.

   SUMMARY: @skynavga to prepare a pull request as per the above
   discussion for further review.

   github-bot, end topic

Use of PNG image format. ttml2#694

   github: [29]https://github.com/w3c/ttml2/issues/694

     [29] https://github.com/w3c/ttml2/issues/694

   Glenn: For this and #605 Mike has suggested that we define
   features, which I'm happy to
   ... do but I don't know exactly to reference.
   ... I can't do anything without a proposal for a format
   definition document.
   ... Like the ISO definition and reference number etc. I need
   guidance.

   Pierre: You can use the definitions from IMSC1.

   Glenn: Did you define a feature there?

   Pierre: No, but if it had one then we could add it to IMSC.

   Glenn: Is there a PNG HDR ref?

   Pierre: There's the document we published, which references the
   PNG format.

   Glenn: I suggest incorporating the reference from the note and
   informatively referencing
   ... the note as additional information.

   Pierre: Both the note and IMSC refer to the same W3C PNG
   document.

   Glenn: Okay. That's fine.
   ... I have enough info to proceed.

   SUMMARY: @skynavga to proceed with the new information provided
   above.

   github-bot, end topic

Use of HDR images. ttml2#695

   github: [30]https://github.com/w3c/ttml2/issues/695

     [30] https://github.com/w3c/ttml2/issues/695

   SUMMARY: As per
   [31]https://github.com/w3c/ttml2/issues/694#issuecomment-389905
   466

     [31] https://github.com/w3c/ttml2/issues/694#issuecomment-389905466

   github-bot, end topic

Negative feature designators aren't future proof. ttml2#697

   Nigel: (summarises issue as relating to feature comparison
   arithmetic and future proofing_
   ... possible solutions are to whitelist the syntax and
   semantics which I understand is a lot
   ... of work, or to reference the original definition from
   TTML1. I wanted input from the group
   ... because that pushes the effort onto the reader of the spec
   to understand the diff between
   ... TTML1 and TTML2.

   Glenn: There's a separate related issue to reference TTML1
   feature designator definitions from
   ... feature designators that are carried forward into TTML2.
   ... That bumps the definition back to TTML1 that is already in
   the wild.
   ... The second part is the request to specify attributes and
   elements and even listing values.
   ... That would be a complicated process and prone to error and
   ambiguity because you either
   ... have to repeat the definition or you have to summarise or
   paraphrase it which is a
   ... reduction that might result in ambiguity. It also breaks
   the "don't say it twice" rule. While
   ... we do that in some places already, e.g. adding rw and rh
   units to length, which are very
   ... simple, but for larger features like ruby-reserve or the
   tts:ruby attribute it would be
   ... impractical to paraphrase the meaning of that so we
   followed the formula from TTML1.
   ... We don't call out all the sub-features of the attribute.
   ... While I admit we already have some features with a
   whitelist those are the exception
   ... rather than the rule and I don't want to set a rule.

   Pierre: The large number of features that would be affected can
   not be a reason not to do it
   ... because the obvious answer is to get rid of features that
   nobody cares about. Or we can
   ... do it.

   Glenn: We'll be here until next year.

   Pierre: Or we can let Nigel do it. The large number of features
   cannot be an excuse.

   Glenn: The review work would be just as much!
   ... You've approached this from a generic speculative
   perspective. I would rather tackle
   ... this from a bottom-up perspective where we fix individual
   problems where we find them,
   ... and if we can synthesise a rule then we can apply that.

   Nigel: I think the counter to this is to mark feature
   combination as at risk.

   Glenn: That's reasonable. Do we have a feature for that? I am
   thinking we don't.

   Pierre: Does your comment Nigel apply to those features worded
   as "Notwithstanding the
   ... above support for xyz is not implied"?

   Nigel: I haven't thought about that - the examples I was
   thinking of were listed in the issue,
   ... such as `#textEmphasis-no-color`.

   Glenn: Nigel you contend that "all defined values" is
   open-ended but I would contend that
   ... it is not, in that the defined values are those in this
   version of the specification.

   Nigel: I suppose that's logical and we're referring to TTML1
   for TTML1 defined features.
   ... Logically we could future-proof ourselves against later
   versions by specifying TTML2
   ... for newly introduced features also.

   Glenn: We need concrete examples to solve here.

   Nigel: Going back to the main question I wanted to ask the
   group: is it okay to refer the
   ... reader back to TTML1 for features defined there? Any
   objections to doing that?

   group: [silence]

   Nigel: I'll take that as assent to adopt that approach then.

   RESOLUTION: Mark TTML1 features as relating to the TTML1
   definition

   SUMMARY: This resolution is logged as #763

   github-bot, end topic

Rewrite version 1 features as references to [TTML1]. ttml2#763

   github: [32]https://github.com/w3c/ttml2/issues/763

     [32] https://github.com/w3c/ttml2/issues/763

   RESOLUTION: Do this, as agreed in today's meeting.

   github-bot, end topic

Disallow `<length> left` and `<length> right` syntax. ttml2#711

   github: [33]https://github.com/w3c/ttml2/issues/711

     [33] https://github.com/w3c/ttml2/issues/711

   Glenn: The open question is whether to adopt the CSS
   restriction on position. My contention
   ... is that there is no ambiguity.
   ... Do we want to remove the ability to have vertical first, so
   you could not say "top left".
   ... When both are keywords, English conventions argues for
   allowing "top left" and "bottom right"
   ... but you want to make `top <length>` not permitted?

   Pierre: Yes

   Glenn: Because in that case the second position length is
   horizontal not vertical and you
   ... want to resolve length as horizontal first and vertical
   second.

   Pierre: Yes

   Glenn: I can implement the CSS restriction if that's what the
   group wants. I think I
   ... implemented this and didn't have any difficulty. I can go
   along with that for CSS consistency.

   Nigel: I would second that, for CSS consistency. Any objection
   to applying the CSS rule here?

   group: [no objection]

   RESOLUTION: Adopt the CSS constraints on ordering

Definition of implicit inline region ttml2#727

   github: [34]https://github.com/w3c/ttml2/issues/727

     [34] https://github.com/w3c/ttml2/issues/727

   Nigel: The discussion point here relates to the pull request
   #731. We don't have implied
   ... inline regions at all anymore, so the feature designator is
   wrong.

   Glenn: I guess I could change #region-inline-explicit to
   #region-inline.
   ... I would prefer `#region-animation-implicit` for the other
   one.

   Nigel: OK

   Glenn: I'll address the wording too.

   Nigel: Okay, thank you.

   SUMMARY: @skynavga to update the pull request as per the
   discussion above.

   github-bot, end topic

Featurize offset times w/o metric. ttml2#752

   github: [35]https://github.com/w3c/ttml2/issues/752

     [35] https://github.com/w3c/ttml2/issues/752

   Glenn: I'll prepare a pull request to bring this back to TTML1
   syntax by removing the
   ... optionality of metric. Then we can hold off on review until
   the pull request stage.

   Nigel: Sounds good.

   SUMMARY: @skynavga to prepare a pull request to remove the
   optionality of metric on time expressions

   github-bot, end topic

Meeting close

   Nigel: Thanks everyone. Reminder: no meeting next week.
   [adjourns meeting]

Summary of Action Items

Summary of Resolutions

    1. [36]Move the tta: namespace style attributes into §10.2
    2. [37]Add a note as per the comments in the issue.
    3. [38]Mark TTML1 features as relating to the TTML1 definition
    4. [39]Do this, as agreed in today's meeting.
    5. [40]Adopt the CSS constraints on ordering

   [End of minutes]
     __________________________________________________________


    Minutes formatted by David Booth's [41]scribe.perl version
    1.152 ([42]CVS log)
    $Date: 2018/05/17 16:42:55 $

     [41] http://dev.w3.org/cvsweb/~checkout~/2002/scribe/scribedoc.htm
     [42] http://dev.w3.org/cvsweb/2002/scribe/

Received on Thursday, 17 May 2018 16:44:15 UTC