See also: IRC log
<scribe> scribe: nigel
nigel: For today we have some
TTML issues and hopefully we will have r12a joining us
... shortly. For IMSC I think we need to record publication of
the WR and think briefly
... about the next steps. I will leave that until later in the
meeting. Any other business, or
... constraints?
Thierry: Someone from VideoLan
sent me a few messages yesterday, and he said that the
... IMSC spec has a lot of issues, not only considering the two
additional features, but also
... the Rec. He said that the test suite is not good enough, so
I have asked him to raise
... those questions to the mailing list.
... So he should send something.
Pierre: Jean-Baptiste?
Thierry: Yes
Pierre: In the context of imscjs
I've created a more thorough test suite based on the TTML
... test suite (with bug fixes and made compliant with IMSC1)
and the IRT test suite, and I
... have slowly been adding additional tests as people bring up
additional test cases. The
... suite consists of the TTML documents and some intermediate
documents and output
... images. I can point to that, and my goal is to submit that
back to W3C as a complete
... IMSC test suite. Just time has been stopping me from doing
that. There are a couple of
... bugs and some tests to add and then it should be
good.
... Please let me know if there are any obstacles to doing
that.
Thierry: It is probably not clear that the goal of the W3C test suite is for interop not for
<pal> https://github.com/sandflow/imscJS/tree/master/src/test/resources/reference-files
Thierry: product testing. I'm not
sure how we explain that, but maybe I could look at it
and
... add some words to explain that. It is good for our purposes
not for implementors.
nigel: However a test suite that is good for implementors will suffice for our purposes too?
Thierry: yes.
Nigel: Just closing off on this meeting, any other points to raise?
group: [silence]
Nigel: Just a reminder, in 2
weeks on 6th April I will be unable to join the meeting so
if
... anyone wants to step in to Chair, please let me know.
<r12a> https://github.com/w3c/ttml2/issues/277
r12a: If we had agreement on this then it might make things easier to discuss in other areas.
Dae: Link to the CSS document please?
<r12a> https://drafts.csswg.org/css-writing-modes-3/#block-flow
r12a: The reason I raised the
issue in the first place is because TTML says it is
following
... CSS but CSS has recently changed before going to CR. The
change was that instead of
... using the vertical-lr and vertical-rl and then fiddling
with text-orientation to make
... latin text run up the page for example there are two new
values for writing-mode, which
... are sideways-rl and sideways-lr and they are specifically
for horizontal text in a vertical
... arrangement, e.g. for text running alongside a table these
would be useful. If you had
... some latin text running up the page and you text-align it
to the end of the line, in the
... old way the text would drop to the bottom, but with
sideways-lr the text rises to the top
... as you would expect. It also makes it easier for line
breaking and the default orientation
... for characters is much easier to handle. This makes things
much easier.
... This leaves text-orientation only applicable to vertical-*
writing modes, which are what
... you would use for Chinese, Japanese and Mongolian most of
the time. They determine
... the default orientation of characters according to Unicode
TR.
... So the text-orientation property becomes a way of
influencing what's happening in
... those vertical- modes, so for text to stand upright you
would apply text-orientation.
... But normally you would use the sideways- values.
Glenn: Right now the writingMode
property in TTML is based on the XSL-FO spec rather
... than CSS.
Nigel: Doesn't it fall through to CSS?
Glenn: No, and certainly not to
the current version. Actually it enumerates all the
values.
... At the time there was no writing mode in CSS.
... Also, since this is based on TTML1 we don't have a lot of
room for changes there because
... it is already deployed in the field. The second point is
that textOrientation right now
... has a default of "mixed" so horizontal scripts would be in
sideways mode. So right now
... if you use tbrl or tblr with Latin text it would put it in
sideways mode so you would end
... up getting what is apparently (I haven't read the details
of the new CSS specs) the same
... as the sideways mode, and you don't have to specify that
since it is the default.
... So I think it works exactly like you sais.
... Also the textAlign default is start which would mean that
if it were bidi Arabic it would
... put it at the bottom in a vertical mode.
... Again it would work correctly for both lr and rl paragraphs
and would derive the textAlign
... from that so Latin script in vertical lr would be aligned
at the top, and Arabic at the bottom,
... so it would get the same result. So I don't see any
particular advantage in pulling in those
... values from CSS at this point given the functional
similarities.
r12a: I suspected there would be
some legacy here so I put in the issue that perhaps one
... issue would be to keep the legacy and if you want to
introduce the new sideways modes.
Dae: One question about the CSS
document. It says the sideways values link to level 4 but
... all those links are broken.
r12a: I thought it said they are at risk.
Dae: The newer editor's version
seems to defer them. I'm not sure which version is newer.
... A general to the question: how do we deal with values that
are at risk or being deferred.
Glenn: textOrientation is a new
property in TTML2 but writingMode was there previously
... so we have to work around that a bit. The current
definition is that the default initial
... value is "mixed" which is defined as [reads definition from
TTML2].
https://w3c.github.io/ttml2/spec/ttml2.html#style-attribute-textOrientation
Glenn: So you don't need to use
sideways because you get the same default behaviour.
... The only time you would need to use sideways is... hmm,
that raises in my mind the question.
... I guess if you had set the default to upright and you
wanted to override with sideways
... you might do that explicitly.
r12a: The mixed default is
perfect. It says "glyphs ... are set 90º clockwise" which is
fine,
... but then it does not clarify in my mind whether if the word
"Hello" would have H above
... or below e - it sounds like only the glyphs are changed so
you might end up with olleH
... in the text. That was a stumbling issue in CSS and why they
moved to the new model.
... The other thing is "Hello world" in English and
sideways-left is applied and then
... textAlign is applied then there is the question of which
way does the text move if it is
... smaller than the box? That's not a glyph thing, it's a box
thing if you like. Not just a
... question of turning the glyphs around but applying a
different mechanism to the way
... the box is working.
Glenn: I think you're talking
about the baseline and referring to the fact that in
vertical
... scripts you will sit on a centre baseline whereas a roman
script will have a non-centred
... Latin baseline?
r12a: No I'm referring to the
distinction between twiddling the orientation of the glyphs
and
... changing the line direction. The TTML2 spec only seems to
talk about changing the glyphs
... individually not the group of glyphs.
Glenn: So if you had
sideways-left then you would expect Hello to have H on the
bottom and
... ascenders pointing right?
r12a: The worry is that might not
be the case, and that might not be the case in CSS.
... You might still read H e l l o down the page and the
descenders would point to the left
... instead of the right, in sideways left mode.
Glenn: I guess I did not say
"clockwise" or "anticlockwise" relative to what. I meant
relative
... to an upright position. That would mean counterclockwise
90º would put the ascender
... to the left.
r12a: Right.
Glenn: In both cases I don't see
anything about if it affects the entire line box as well.
In
... other words model as though setting horizontal text and
rotate the line instead of the
... glyphs. OK that's a fair point.
r12a: When the CSS folk realised
that they switched to new values for writing-mode
instead,
... because that affects alignment, position of characters in
the line box, etc. so it turned
... out to be a convenient way of looking at the world.
Specifically for use with what are
... normally horizontal scripts - there are plenty of
situations where you want them to run
... up or down the page rather than horizontally.
Glenn: That's fair. In TTML1 we
had an example of a latin script in a vertical mode and
we
... used upright glyphs in the example which raises the point
of if it is valid to change to
... mixed for TTML2 and have "mixed" cause rotation, which
would invalidate that example,
... even though the example was informative. That could be an
issue we should document
... and talk about more.
... I understand the comments now, Richard, thanks for those. I
want to give it some
... more consideration, but I think you also understand some of
what TTML2 has done at this
... point. There are also some existing implementations of the
new features that have been
... deployed so we have to be a little sensitive about those
too.
r12a: One more thing: these
sideways values in CSS are best thought of as a way to
... rotate the box rather than the character glyphs. It helps
with understanding what is
... going on.
... That's not necessarily what happens, but it's helpful for
thinking about it.
Glenn: There may be some value in
that to do with the issue of whether you start at the
... top or the bottom. I need to cognate on that a bit more.
The rotation of glyphs though
... is still a relevant point because fonts have rotated
variants and in the case that a font
... does not have rotated variants it may perform the rotation
manually on a glyph by glyph
... basis. This comes up in asian scripts more than in roman
scripts. But the same rotation
... dimension applies to any font.
r12a: CSS does still have a sideways value for text-orientation for that kind of scenario.
Glenn: I had a conversation with
@fantasai about this a few years ago, after which I think
... the sideways versions of text orientation were taken out.
We had to rely on the state
... of the CSS definitions at that time.
Dae: Do we know when Level 4 is
coming out? Are we confident that the definition of
... sideways won't change?
r12a: I don't know the a-z of
that - I know she read this and agreed with what I was
... recommending. Firefox has implemented sideways-* already,
so I am hoping that we
... would get Chrome to support those also, so I will follow up
on the status of that.
Glenn: Dae I think we cannot make
any schedule dependent on a CSS document appearing
... at Rec, so keep that in mind.
Pierre: Flipping the question around: can the CSS document reach Rec? And by what time?
r12a: I know they have been
talking this week about fast-tracking some specs
including
... Writing Modes so I am hoping it will reach Rec.
Pierre: It is hard for this group
to follow a CSS spec without a clear schedule, because
the
... risk is high.
r12a: That is something to talk
to CSS about - it would help them I think.
... Like Glenn said, it would be great for users as well if CSS
and TTML work in similar ways
... because they would not need two ways to think about things.
There is the possibility
... of moving ahead with this if they make sense. There are
lots of examples of that in
... TTML2 for example with Ruby, alignment and so on.
Glenn: I have brought some of those up in the past with Elika and Koji.
Nigel: Let's move on to the next issue then:
https://github.com/w3c/ttml2/issues/240
Nigel: Pierre I think last week you thought we may be able to conclude on this?
Pierre: My conclusion from the
thread is that there is a mapping for every option in CSS
... to every option in TTML and the discussion is about the
best default.
r12a: That's my understanding as
well. There are three things that have been coming into
... the conversation. They are to do with things that sit
alongside the vertical line.
... One things is text emphasis marks, another is Ruby, and
another is "lines". Actually
... the position of those things may be different depending on
which thing they are.
... There is a Ruby issue for #240 but that might not be
appropriate until we've decided
... what to do with textEmphasis. Then for me the question is
what is the default specifically
... for tblr - I'm not worried about tbrl.
Pierre: The default in TTML2 for
some writing mode might not be or might be what
... someone expects depending on language. The challenge is
changing the value of
... the default based on the value of something else makes the
processing algorithms very
... complex. That would be an implementation burden. If the
issue is simply one of defaults
... we might have to accept what TTML is today.
Glenn: I concur with Pierre. We
need to have a default for each property, that's a given.
... There's no argument about whether "before" should be the
effective default for tbrl or
... any of the horizontal modes. The only question is if it
would be appropriate if you were
... using Mongolian, which apparently puts text emphasis on the
after edge for tblr writing mode.
... Now Mongolian can also be written in tbrl where you would
want it to be on the before
... side (the right side).
r12a: That's not true - you never write Mongolian tbrl.
Glenn: I have seen a number of
manuscripts at Harvard that were directly from the
imperial
... language in Beijing and those were all in tbrl.
r12a: Were they Chinese documents? In that case you would see tbrl.
Glenn: They were pure Mongolian
with occasional Chinese characters in. They were created
... using the same writing practice as normal in Chinese with a
scroll that gets unwound to
... the left and the right roll gets wound in. In Mongolian
they shift the scrolls 90º counter-clockwise
... so in those scrolls they look tbrl in pure Mongolian. I've
never seen a tblr Mongolian
... document in my own research but I haven't looked at modern
Mongolian other than what
... I have seen in Chinese word processors in Inner Mongolia. I
don't recall looking at text
<r12a> https://github.com/w3c/type-samples/issues/55
Glenn: emphasis in those contexts
though.
... For me it's like Pierre said, adding a variation in how to
derive the default behaviour
... based on potentially a resolution of xml:lang would
certainly complicate things and
... from an implementation perspective it would be much easier
to apply "before" to all modes
... and then if one wants to do something different use the
initial element in TTML.
... There's no argument about what Richard says.
r12a: I just put a link into IRC
by the way with some Mongolian text. I have been working
... with Mongolian experts over the last 2 years and seen a lot
of Mongolian text. This example
... was sent to me a few days ago, with "underlining" that
appears on the right hand side.
... All of the text I have seen goes tblr. Let me clarify that
I'm not asking for a decision to
... be made based on the language. I am asking for the default
to be dependent on whether
... we have tblr or tbrl as the value, which is what CSS does.
Also to clarify that the requirements
... for Mongolian as I understand them from talking to
Mongolian folks is that the same side
... is used for the emphasis as Japanese, which means that for
horizontal the line appears
... on the other side, the same as Japanese, on the right for
vertical or underneath for
... horizontal. CSS has the same default for Mongolian as for
Japanese for underlining.
... For text emphasis I don't know if they actually do dots or
sesame seeds or whatever in
... Mongolian. The picture in the link uses a line for
emphasis.
... For Ruby, it is clear that it is always on the right hand
side. I asked three Chinese experts
... this morning how this works. In Chinese the line is used
for identifying names of people,
... book titles etc and it goes on the left hand side unlike
Japanese or Mongolian. The text
... emphasis may go on the left hand side but typically would
go on the opposite side from
... the line used for book titles etc. They said you don't see
much vertical subtitling in Chinese
... at all, and where it is used the line progression is right
to left rather than left to right.
... In Mongolian you would expect the line progression to be
left to right.
... My worry is that in most usage these days, tblr is for
Mongolian, but if we are using
... before or after then for every Mongolian subtitle you have
to make sure you get things
... on the correct side. It is not quite so straightforward
because in some cases things move.
... As I understand it in Mongolian the line goes on the left
and the Ruby goes on the right.
... It seems unfair for these, the main users of this
orientation, to have to set a different
... default. It is not based on the language but on the
language model.
Glenn: In a TTML2 document you
just have to put one element up in the head of the
... document so it is not much of a burden to do that and it
makes processing more
... consistent and reduces the implementation special cases to
worry about.
Pierre: Can we add a note pointing out Richard's observations?
Glenn: That would be quite appropriate, and to advise on what to do. Would that work?
r12a: That would help, yes. I
agree it's only one change to make. The bigger concern is
that
... it does not address the sensitivities of people who may
think there is a western or a
... Chinese bias being introduced.
Glenn: I understand that very
well!
... I am also sensitive to implementers, and testing. It seems
like an awful lot of work
... to change the behaviour to make the default sensitive to
the writing mode and also as
... I have pointed out in the case of Chinese I have on many
occasions seen tblr including
... Ruby on the left and text emphasis on the left.
r12a: By the way text emphasis
does typically occur on the left hand side but the Ruby
... would be very unusual especially if you are using Bopomofo
so that's a difference between
... Japanese and Chinese.
Glenn: One question: at least in Mongolian proper ("outer Mongolian") Cyrillic was the
<r12a> https://github.com/w3c/type-samples/issues/56
Glenn: primary script used for
many years, has there been much movement to going back to
... the primary script?
r12a: There is a strong desire to
go back to the original script and I can produce lots of
... examples quite easily.
... I have seen books and booklets in Mongolian, I don't know
about newspapers.
Glenn: It would be a useful point
of note to hear what someone in a standards body in
... the mainland's position would be because they have an
official use of the script for
... "inner Mongolia" on the Chinese side.
r12a: Most of my conversation has been with them actually.
Nigel: So one possibility is to
note the option to use the initial element and another is
... to make the default writing mode dependent - is that what
CSS uses?
r12a: Yes, in CSS the terms are "over" and "under" and they are writing mode dependent.
Glenn: We could add over and
under, either now or at some point in the future. My
... preference is to add a note and not add over or under at
this time.
Nigel: Does that mean that the mapping to CSS will be harder if we do not add over and under now?
Glenn: I'd have to think... If we
add over and under in the future, how would that impact
... the default now? The default now is based on auto which
already has a quirk based on
... the number of lines - it maps to "outside" for two lines
which is definitely not in CSS,
... and that came about from subtitle usage in Japan, not
considering Mongolian usage.
... If it is not two lines then it maps to "before". Even with
what we have now it is not unlikely
... that you would end up specifying a different behaviour e.g.
"before" if you don't want
... this outside behaviour. If we added over and under in the
future that would not change
... the default setting so you would still end up having to
specify e.g. over instead of before
... for the CSS behaviour now.
Pierre: My 0th order concern that
there is nothing missing in TTML is met.
... The choice of writing-mode dependent default is a departure
from TTML so more
... error prone. Adding over and under seems a middle ground.
Adding a note seems
... worthwhile regardless. The main point is that CSS diverged
from XSL and TTML so we
... have to cover those gaps as we move along.
Nigel: And you did not have a concern about mapping to CSS?
Pierre: There is a static mapping, so no.
Nigel: The group's view seems to
be to go with a Note primarily, and that could be enough
... to address the sensitivies of Mongolian readers and
writers?
r12a: If you add the note then I will seek review of it.
Pierre: It is more than that, the
use of initial is the ultimate way to address cultural or
... personal preferences.
Glenn: I agree. The question should be "can you live with specifying initial?"
Nigel: Ok I have added a note to
the issue about this.
... Now lets move on to:
https://github.com/w3c/ttml2/issues/254
Nigel: Have we actually just discussed this?
r12a: The default preference may
be different for Ruby and lines and text emphasis. In
... Chinese the most likely thing is text emphasis on the left
hand side regardless of lr or rl
... but Ruby would be on the right hand side. I'm not sure that
the answer for #240 would
... provide the answer for this but I think that the final
result is going to be the same.
Nigel: Is this the same as the discussion we just had?
r12a: It is, essentially the same and the thing that lead to #240.
Nigel: Ok I've added a note to
the issue on that.
... Moving on to #253
https://github.com/w3c/ttml2/issues/253
Nigel: My question is do we have enough data to agree if we need to support bopofo Ruby?
r12a: We still need data about if
bopomofo is used in subtitles for Chinese and I was not
... able to get an answer to this yet when I was talking to
people from Taiwan. At a minimum
... we should not close out the possibility of adding it later
if the need for it becomes stronger.
Glenn: I don't see why we could
not add support for character based Ruby in the future,
... either in IMSC or in TTML2. I've been watching [scribe
missed audio] ... I wouldn't rule
... out the use for pedagogical purposes but I don't have any
data points there.
r12a: It is very much a Taiwanese thing, not used in mainland China.
Nigel: Are we expecting a response to the question on the issue any time soon?
r12a: I will take an action for myself to prompt him on that.
Dae: I think this is a primary
candidate for TTML.next - our major subtitle authors
... based in China never use bopomofo.
r12a: One possibility is that
again you could put a note in saying we know we do not
... support bopomofo ruby at the moment and even add a request
for contact if anyone
... thinks it is important. At least this shows we have not
been ignorant of it.
Glenn: That's a good point.
Nigel: Okay I have added a
comment on the note explaining our default position on
this
... in the absence of further data points.
https://github.com/w3c/ttml2/issues/253#issuecomment-288760842
Nigel: Thank you very much for joining us today Richard!
Glenn: Thanks for your input Richard.
r12a: Thanks, bye!
Nigel: In the time remaining can we very quickly cover the open pull request on TTML1?
https://github.com/w3c/ttml1/pull/233
Glenn: I believe this is
something Andreas posted. I need to coordinate it with TTML2
actions
... around the same text and make sure we are consistent and we
have dealt with the errata
... aspects of it.
Andreas: I think this makes
sense. It has been open for 1.5 months so I think we
should
... make progress on this to clear it from the table.
Nigel: Looking at the history this is an action for Glenn then.
Glenn: My only concern there is I
will probably edit it to shorten it quite considerably.
It
... seems overly detailed as a note because I don't think we
want to dive into some of the
... language around XSL-FO semantics like large allocation
rectangle etc so I would like to
... say the same thing paraphrased to make it shorter.
Andreas: Then this means it is not actually accepted.
Nigel: We'll have to review the edited version.
Pierre: As a matter of process it
is weird to have a pull request open for so long. What is
... the deadline for getting this addressed?
... It is weird for the group to agree the essence of the pull
request and then have nothing happen.
Glenn: We need to have the
language the same in TTML2 and TTML1 so there is a
dependency.
... We haven't prioritised any updates on TTML1; TTML2 has a
higher priority.
Nigel: We generated a big list of discussed and agreed issues for TTML1 in London.
Glenn: We don't have a schedule for a third edition.
Nigel: We don't have an agreed date for it.
Glenn: As soon as I have it addressed in TTML2 I will backfill into TTML1.
Andreas: I would like to support
what Pierre said - we are quite fast in merging pull
requests
... in TTML2 and I would like to do the same in TTML1 also.
Pierre: Can we prioritise this one?
Glenn: Sure I will put it to the top of my list and do it today, in TTML2 and TTML1.
Nigel: OK thanks.
Nigel: Thank you to Pierre and Thierry for publishing the WD for WR.
https://www.w3.org/TR/2017/WD-ttml-imsc1.0.1-20170322/
Nigel: That was published
yesterday, and Thierry posted many messages to other W3C
... groups requesting wide or horizontal review.
... But not all groups - Thierry you need some input from the
group?
Thierry: I'm missing privacy and security and TAG.
Nigel: Please could you do a
first draft of the privacy and security questionnaire for
the
... group to review?
Thierry: Yes I will.
Nigel: Okay, adding the issue...
https://github.com/w3c/imsc/issues/222
Nigel: Thank you, and I have an
action as previously agreed to send outgoing liaisons to
... the usual groups based on the text that we agreed a while
back. I think only the review
... response date needs to be edited.
... I will be able to do that probably on Monday if I can't
find time to do it any sooner.
Thierry: So we might receive some
comments - I propose to track those and then if we
... need to come up with a document at the end then ...
Pierre: Can we do it through GitHub with a Wide Review comment tag on the issue?
Thierry: Do you want to use the Last Call tool or do something different?
Pierre: I assume we should track it on GitHub. Is there a downside to that?
Thierry: No I don't, at some point we need a disposition of comments document.
Pierre: I assumed we could use
labels on GitHub and then generate say a PDF to be the
... document.
Thierry: We need the comment, our resolution, and agreement from the commenter.
Nigel: I would suggest recording
the issues in GitHub and also using something like the
... Last Call tool to generate the documentation that we
need.
Thierry: I am happy to maintain both, and I don't anticipate hundreds of comments.
Pierre: What is important is that
every change we make to the document is tied to an
... issue in GitHub.
... If you prefer to use the LC tool we can probably create a
bunch of GitHub issues at the
... end and link back to the tool comment if we can, but a lot
of folk look at GitHub to see
... what issues are open. My intuition is to enter the issues
on GitHub to be transparent.
Thierry: We'll start like that
and if needed and there are a lot of comments I will use
the
... tool. Even with 10 comments it is faster to do it
manually.
Nigel: Great I think that is everything for today and we have hit time, so thank you everyone. [adjourns meeting]