Timed Text Working Group Teleconference

07 Apr 2016

See also: IRC log


Nigel, Glenn, Pierre, plh


This Meeting

<scribe> scribe: Nigel

nigel: For today, we have IMSC -> Rec, Charter, and TTML2 stuff. AOB?
... Plus actions of course.

group: No AOB.

Action Items


<trackbot> action-458 -- Glenn Adams to Create issue re: line height calculation for inline -- due 2016-04-07 -- OPEN

<trackbot> http://www.w3.org/AudioVideo/TT/tracker/actions/458

glenn: I'll add that issue after this meeting.


<trackbot> action-459 -- Thierry Michel to Process the PR at https://github.com/w3c/tt-profile-registry/pull/3 and post for review to the reflector -- due 2016-04-07 -- OPEN

<trackbot> http://www.w3.org/AudioVideo/TT/tracker/actions/459

pierre: Thierry kindly offered to apply the pull request and then send the document to
... the reflector for a 2 week review so we can decide what to do with it.

plh: Is this something urgent?

pierre: I don't think so - it's just that Thierry volunteered. People noted that they haven't
... reviewed the document for a while, and would do so after merging the PR. I think
... Thierry was keen to publish it as a WG Note, so he took the action.

nigel: Thierry isn't going to be able to do that for a couple of weeks so I'll extend the date on that action.


plh: This has been waiting on me to progress with W3M. I'm hoping to do that later today
... or tomorrow, and for the Charter to go to AC next week. It's been presented already,
... so I don't expect any objections - just minor tweaks.
... Today or tomorrow I'll create a Pull Request for the changes.

nigel: And in the meantime the current charter has been extended by 2 months?

plh: Correct, for the AC review. You guys should proceed as normal, as you are doing.


plh: The deadline for the AC review is tomorrow, so assuming we don't have any
... objections by then we may be in a position to publish on 14th April, or if not then
... the 19th. I'll try to go for the 14th. If you would like 2-3 days notice to confirm that
... then I can send you 2-3 days notice.

nigel: Yes please!

plh: In that case if I can't get approval by Tuesday to move it forward then I'll push it
... back to the 19th.

pierre: Coralie and the comm team may appreciate some coordination there.

plh: Are we issuing a press release?

pierre: Yes, I think we are planning to.

plh: In that case I will coordinate with them.

pierre: The goal is to have a draft tomorrow - I've been talking to Karen about it.
... I owe her today some technical bullet points about IMSC 1, which I plan to copy others
... on for review. She had taken responsibility to take on the broader PR exercise, I understand.

plh: Okay, I'll work with her on her timeline because I'm guessing that would be the
... bottleneck. I'll check on her timeline and CC you guys. On the AC review side unless
... there are any unexpected objections.
... Did we receive any comments on IMSC 1 since PR, that we could share with the Director?

pierre: I'm not aware of any.

plh: If you see anything then let me know before Monday. I'll go and check myself.

pierre: There are no issues against IMSC 1 PR on github.

plh: Okay, should be smooth, aside from the Comm team's timeline if we want to target Tuesday next week.


nigel: Are you aware of the issues we've had with the new stylesheet plh?

plh: I'm not aware of any specific issues you guys are facing.

glenn: It's the use of the XMLSpec stylesheet to generate the HTML. The format that it
... generates for the table of contents is incompatible with the new sidebar layout of the
... table of contents, so I need to do a fairly significant revision of the ToC generation
... rules in that stylesheet. It's involved and will take some time but it's feasible.
... It's a non-trivial change. Others that are using XMLSpec (few probably) will have
... similar problems.

plh: Have you checked with spec-prod to see if anyone else has had this problem and
... resolved it?

glenn: Well no - it's pretty crufty, and we have over time considerably customised it
... for TTML in any case, so just dropping another group's work into the TTML transform
... would probably not be much easier than doing it ourselves. I'm not sure who else
... is using XMLSpec - perhaps some of the XML activity?

plh: Yes, maybe. I don't think there's much I can do to help at the moment.

glenn: On the lineheight discussion I'm preparing some additional graphical materials
... to help the discussion process since it's difficult to discuss without them.
... One thing I found: in CSS2.2, which is the next to be released official version of
... CSS, under ยง10 I believe, there's an interesting comment I've found.


scribe: "Note. CSS 2.2 does not define what the content area of an inline box is (see 10.6.1 above) and thus different UAs may draw the backgrounds and borders in different places."

glenn: So it would appear that the CSS group has acknowledged the problem we are
... facing in that note, which we should keep in mind.

nigel: Does that seem to make the problem worse or give us a degree of freedom?

glenn: The latter. It justifies us in attempting to nail it down more. We have to be
... cognisant that if we're translating to HTML/CSS then it might not be possible to
... rely on CSS alone to do what we want it to do.

plh: Do we have a list of what different UAs do?

glenn: Very good question. All of the browsers I've looked at so far take the text altitude,
... the text depth and the text line gap metric for the font and use that as the height of
... the content rectangle area (horizontal text). They have basically reverse engineered
... each other and fine tuned their implementations to match each other aside from the specifications.

plh: Would that be okay for TTML2?

glenn: I've proposed contentBpd and the default "auto" value I propose to make equivalent
... to what the browsers implement right now, with another setting that would create the
... effect that we want in TTML2 which is different from that default behaviour.
... If in a TTML2 document mapped to HTML/CSS a newly defined value were used then
... the auto value would map but others would not.

nigel: That would not stop any implementation though - you can use Javascript getClientRects for example.

glenn: Sure and you could use SVG.

plh: We should consider raising an issue on CSS.

glenn: I've had a private conversation with Bert Bos on this so I'm going to suggest that
... when we get to the right point that we inform the CSS group of our solution.
... 10.6.1 of CSS 2.2 says "Note: level 3 of CSS will probably include a property to select which measure of the font is used for the content height."
... Right now CSS 3 doesn't say anything about it though - I don't know if anyone is working on that.

nigel: And in the meantime our approach for TTML1 is to add an errata. Is that still progressing?

glenn: Yes.
... We could move on to the <initial> element.

nigel: Yes, let's.

glenn: I don't think there's any problem to solve. The initial element has been defined and
... is used out in the field. I don't see any issue with it other. On the issue of whether it
... is required, it's deliberate, because the the region root inheritance only applies to
... inheritable properties. If you wanted to use a different value for that then you're out of
... luck. Also some of the initial values that we define are implementation dependent,
... including tts:color and tts:fontFamily. The mapping of "default" to a platform font
... is undefined, so that works out to be the same situation as tts:color.
... There are cases where there are non-inheritable properties where you want to use
... something different from what's in the spec, and inheritable ones. There are solid
... use cases for this. As has been pointed out you could always explicitly state the styles
... but this is a known problem with TTML and it in part was originally motivated a number
... of years ago when the first SMPTE-TT was being published. In the initial draft they
... had defined a different initial value. The same thing then occurred in EBU-TT. I had
... pointed out in both those instances that it was effectively non-compliant with TTML.
... Those changes got backed out, but it still left the problem of wanting to use a different
... initial value. I think we discussed this maybe 6 years ago and has been on the books
... for a long time. I don't see any point in questioning the current status now.
... Nigel has raised some other interesting things but I don't want to reopen the case of initial right now.
... I also mention that Netflix is making heavy use of initial in TTML2 right now.

nigel: Where are they seeing benefits rather than just specifying style in the normal way?

glenn: I've mentioned color and font already - they have put in place a number of
... production tools that are translating from a number of other formats into TTML2 and
... they are making use of initial there based on the fact that this was in TTML2.

pierre: I think we should answer the question why use initial rather than referential styling?
... Second I think the fact that that feature has been in a FPWD for a year is a data point
... but is not the only one. Most importantly I'd like to understand why referential styling is not being used.

glenn: It can be used.

nigel: It's unclear why one would use initial?

glenn: It's about the efficiency of intermediate representations. The ISD format in TTML2
... does not specify the value of styles that are the same as the initial value.
... Most styles are resolved to their initial values and those are elided for efficiency in terms
... of information content. The same principle applies at the TTML2 level where if
... referential or inherited styles are used then there's no need to specify initial values.
... I've encountered many cases where tools generate explicit styles that happen to match
... the initial values, which is inefficient, bloats the document, requires more bytes to
... be transmitted etc. Ultimately I think the most compelling argument is efficiency.
... As I've pointed out you certainly can use referential styles if you want to.

pierre: People are going to use styles anyway. So it seems like little difference in efficiency
... between using referential styles and initial.

glenn: For that to be true you're forcing the use of referential styling that they may not want to do.
... They may prefer a more efficient way.

pierre: The opposite could be worse though - folk that do not want to use referential
... styling are causing a new feature to be created.

glenn: If you don't want to permit the feature then profile it out. We aren't trying to
... create a single orthogonal set of features here in TTML2.

pierre: My issue is that adding more stuff to TTML2 will make it harder to get it adopted.

nigel: My point in my email was that we need to consider styling in the round and make
... sure that we have a level of coherence and consistency in the specification. If we
... make changes to the styling approach then we may then decide to revisit the initial
... element.

glenn: We had discussion around 2003-4 about more complex styling. We decided to
... postpone it. It may be that we should look at it now. Sure if there's a new proposal
... then we can evaluate it. Right now initial is implemented and deployed and its wasting
... our time discussing it.

nigel: There are two contexts that I think we will need to consider here.
... The first is customisability/personalisation, which is generally accepted as being
... important for accessibility.
... The second is for mapping to HTML/CSS and deferring to downstream style processing
... such as tunnelling class attributes etc.

glenn: One thing we could generate is a new section that makes generic statements
... about the possible approaches to customisation even if it's a placeholder to show how
... we may develop things further.

nigel: +1
... I'm unclear what all the options are but I'm happy to try to generate something there.

glenn: The status quo is that it's player or implementation dependent.

nigel: Clearly, yes. that's one option.

glenn: That's what is done in the browser world today. That would be the minimum it
... should talk about.

nigel: Yes! I'm anxious not to sweep away all the good work in terms of conformance language that's already present.

<scribe> ACTION: nigel Creat an issue to draft a TTML2 section on customisation. [recorded in http://www.w3.org/2016/04/07-tt-minutes.html#action01]

<trackbot> Created ACTION-460 - Creat an issue to draft a ttml2 section on customisation. [on Nigel Megitt - due 2016-04-14].

nigel: We're out of time now, so I'll adjourn. Thanks all [adjourns meeting].

Summary of Action Items

[NEW] ACTION: nigel Creat an issue to draft a TTML2 section on customisation. [recorded in http://www.w3.org/2016/04/07-tt-minutes.html#action01]

Summary of Resolutions

[End of minutes]

Minutes formatted by David Booth's scribe.perl version 1.144 (CVS log)
$Date: 2016/04/07 16:34:10 $