See also: IRC log
<inserted> scribe: nigel
Nigel: Apologies for the late
start today.
... Our agenda today is at https://github.com/w3c/ttwg/issues/6
... [iterates through agenda]. Any other business or particular
points to raise?
Pierre: It might be too late, but
we have this plan for requirements, but have not formally
informed other organisations
... who have cared in the past. I think the announcement was
public so we can point people to it but maybe we ought
... to at least alert SMPTE, EBU. SMPTE meets next week. I
wonder if a really quick liaison might be worth it, later
today
... or tomorrow, saying we're gathering requirements, see this
email for detail. For discussion if there's time.
Nigel: Thanks, that's a really
good point.
... Anything else for the agenda?
group: [silence]
Nigel: I don't think there's anything to discuss here, I did give Glenn the prod he requested.
Glenn: I just got back from travelling and am catching up, I'm afraid I'm behind on this but will push on this next week.
Nigel: Okay.
Nigel: Pierre's comment above
relates to this agenda item.
... I'm happy to draft something lightweight and send it out
tomorrow, it's a good point.
... Organisations who should receive it are, off the top of my
head, SMPTE, EBU. DVB, ATSC, HbbTV, ARIB. Any others?
... I will also check the liaisons page.
Andreas: I just wanted to comment
that my colleague Peter tho Pesch recently added two issues in
the requirements
... repository of the XR Community Group on subtitles,
especially on requirements for 360ยบ videos.
... He will send the link to the issues. This is a good place
to discuss what could also be covered in a future version
TTML.
Nigel: Thanks that sounds good, if you could ask Peter to raise at least one issue on tt-reqs so we can track that please?
Glenn: I'm wondering if it's
worth discussing what are ground rules are for, say, TTML2 2nd
Ed vs a new version of TTML,
... e.g. TTML3 in particular whether or not we are considering
new features in TTML2 2nd Ed or if we are considering
... those only for TTML3. In other words is TTML2 2nd Ed just
bug fixes or are we going to try to slot new features
there.
... We haven't discussed this yet. I don't know if we need to
discuss it right now.
Nigel: A quick response on that:
first let's gather the requirements and then work out how we're
going to deliver them.
... I think this is one of the agenda topics for our face to
face meeting.
... We said we wouldn't work this out immediately.
Glenn: I'm comfortable with
that.
... The only reason I mention it is because as I try to
incorporate requirements into the draft requirements document
I
... need to frame them somehow. I can delay doing that for the
time being.
Nigel: Sounds good.
Nigel: (I haven't updated
github-bot with the tt-reqs repo yet)
... The discussion here is about if we should only add specific
known CSS properties into TTML or one of its profiles or
... if we should do as the issue requests which is to allow for
generic CSS properties to be specified in document
instances.
... The point here is that there's a class of use cases where
waiting for new CSS properties to be added to the spec on
an
... annual cycle isn't feasible.
Pierre: That's a good point. I
think in this case the way to do it is to create attributes in
foreign namespaces so folks
... can experiment to their hearts' content.
... A new feature is going to require a change to
implementations regardless. Using existing CSS vocabulary does
not make
... things easier.
... I'm trying to think of a scenario where importing CSS would
be easier than dealing with specific new attributes.
Nigel: The idea is for a class of
processor that simply passes the CSS properties to a CSS based
implementation without
... having to process them any further, and also to put this
processor functionality behind a feature designator.
... I don't think it is especially helpful for standard
subtitle and caption presentation for example where the use
case is
... much more tied down.
Glenn: A couple of comments. One
is that supporting this feature entails having a CSS processor
and integrating that
... into whatever implementation we're talking about. This
would have for example a fairly big impact on TTPE and
TTV
... for example.
... The other is that in order to make use of any CSS features
from a standardisation perspective we would end up having
... to document each CSS style property that would be used in
this feature that describes how it fits into the TTML
... rendering model. We have based everything conceptually on
the XSL model but it has semantics that have in some ways
... stayed close to CSS. I can certainly see how this might
work for edge scenarios for example the one Nigel brought
up
... in his proposal, like specifying a left and right border
independently of all borders, for example. But let's say you
want
... to support flex display mode via this mechanism. Now all of
a sudden you need to figure out how flex semantics
... fit into TTML formatting. It may work better in some cases
than others and will potentially have interop problems.
... Those are just the concerns that come to the top of my
thinking right now.
Andreas: I actually like Nigel's
proposal, actually everything that brings us closer to the
evolution of CSS and gives us
... a more direct way of using this process in our
standardisation work. I agree we maybe need to play through a
concrete
... example and see how it works but in general I think it's a
worthwhile proposal to check.
... I think that we agreed, at least the majority of the group
agreed, to have a closer mapping to CSS in the future.
... Nobody knows how it could be done because as Glenn says
TTML is not built on CSS but on XSL-FO. It's a step in
the
... right direction. After a concrete example we should
reconsider it.
Glenn: 2 more things.
... The syntactic proposal that is on the table, or something
like it, I think would be quite straightforward to
support.
... The semantic one is the more problematic one. Another point
I didn't mention before is that such a mechanism would
... have to entertain the possibility of users using this
mechanism to specify properties that are already defined in
TTML
... but slightly differently, lineHeight etc. Now you have a
problem what to do. There's a lot of detail to deal with
to
... effectively do this.
Nigel: The order of precedence point is already in the issue.
Pierre: If only the BBC wants to
do this in their own code this doesn't need to be in the
standard.
... There's already an extension mechanism.
Nigel: That's true, my preference
is to standardise but you're right if the BBC is the only
organisation with a use case here
... we can do our own thing.
Glenn: I think it's useful to
continue on this to the extent that we potentially scope out
and spec out a syntactic mechanism.
... In the end, if the WG doesn't decide to include it in some
v.next then if we've defined a reasonable syntactic
mechanism
... then BBC could in fact employ that in their own namespace.
Whatever is done here I would insist on it being in a
... qualified namespace not an unqualified attribute for
example. Adding a bare class attribute would be something I
would
... oppose. I would want to see this being in some namespace
qualified attribute space.
Pierre: One additional point in
response to Andreas. If the long term goal is alignment with
CSS then I'd rather spend
... effort pursuing that. Maybe the long term goal should be
for TTML to use CSS period. If that's the goal let's just
go
... there.
Glenn: We do have a precedent in
W3C for something similar to this, in SVG. It has a set of XML
defined properties just
... like TTML that do not depend on CSS. Then many
implementations added support for CSS stylesheets on top of
SVG.
... Most implementations support user specified stylesheets.
That's a deployed existence proof for doing something
like
... this. I don't know what transitioning to pure CSS means. We
made a design decision in the beginning of this group to
... use XML syntax and not CSS syntax. Short of going back on
that underpinning principle we still have an XML syntax
... for style properties and I don't know how we could move
away from that without a non-backward-compatible revision
... of the current syntax. I fear that if we dive into that
discussion it is going to derail the immediate goal of what BBC
is
... proposing here as well.
Pierre: Glenn, my point is that
BBC's proposal here has implications on semantics that will
force us to solve the bigger
... issue. There is no perfect 1:1 mapping here. It's close,
but applying CSS semantics to TTML is going to be
... straightforward in some cases and not at all in others.
Glenn: I agree.
Pierre: This will force us to solve the deeper issue, of using CSS period.
Andreas: In general I agree with
Pierre I think. Possibly the BBC proposal is a good opportunity
to discuss this question
... and elaborate how we can use CSS more directly in TTML.
Glenn is right that over 10 years ago there was a
decision
... to use XML and base styling on XSL-FO but the situation has
changed a lot since then so we need to face the new
... situation and find an answer for a future-proof model.
Nigel: Thanks everyone for the great discussion, shall we move on from this topic?
Glenn: One last comment. I
haven't looked at what SVG did to accommodate CSS usage but
it's a really important
... precedent given that it mixes XML styling with CSS, which
mirrors TTML in terms of the overlap between properties.
... Perhaps Nigel or somebody promoting this can take a look at
SVG to see if there's any guidance there.
Nigel: Good point, thank you.
Glenn: A quick comment on
requirements in general, I had written in the guidelines about
the requirements that
... people are encouraged to create a marked up document with
more guidance on the issue. I'd ask people raising
... requirements to try to undertake that.
Nigel: Noted, I did intend to do
that.
... On the topic of the meeting, the dates and location are all
confirmed.
Andreas: We discussed this week
the joint f2f meeting and had the idea to kick off the joint
meeting with some
... short presentations, on implementation work, the EBU-TT
Live Toolkit, some requirements from broadcasters about
... contribution of live subtitles to web streams. BBC, Red Bee
Media and IRT are interested in giving short
presentations.
... Is this okay with the TTWG members? If yes then we can
allocate time later. I think it should not be too long,
maybe
... half an hour maximum.
Nigel: Any objections to doing that? Any requests or proposals for additional presentations?
Andreas: I think it would be
interesting to hear if there are requirements from outside
Europe that are not covered, for
... example from north America. That could be interesting if
there is anyone to present it.
Glenn: I know I've mentioned it a
couple of times in the past. One thing people have kicked
around in the past is some
... support for karaoke. That's a possibility. I'm not
volunteering but will try to reach out to those folk.
... We haven't logged an issue yet that records a requirement
in this area. I haven't received any concrete proposal
yet
... but it is possible that one will appear before then.
Nigel: I'm hearing no objections
or proposals for other presentations so right now I believe we
have consensus to go
... ahead with that idea Andreas.
Nigel: Thanks for raising that Andreas, if you want to go into the wiki page and add those in, that would be great.
Andreas: Yes, I will check back with the presenters and then add them.
Nigel: Thank you!
Nigel: I just want to say
something quickly about this. Yesterday there was a special CSS
WG meeting to discuss
... line sizing, whose minutes should be out soon if not
already. In particular I raised the question about whether
ruby
... reserve is included in the line height calculation or not,
and if so, how.
... This got raised as an issue, and the initial view was that
any ruby reserve space should be added to the leading
<glenn> https://lists.w3.org/Archives/Public/www-style/2018Nov/0032.html
Nigel: to achieve the necessary
effect.
... So folk interested in line sizing and ruby reserve might
want to go and take a look at that.
... The issue of background area painting did not arise because
it's orthogonal given that there's already a draft
property
... for setting the background height to extend to the line
edges.
Glenn: What is that property, for the minutes?
Nigel: The "stretch" value extends the background to the line box edges.
Nigel: We don't have David on the
call, but let's do what we can.
... The first thing to do (which everyone else did before I
came late to the call) is to welcome Gary who is the new
Editor
... of the WebVTT spec.
Gary: There hasn't been much
activity but I heard that there needs to be work done to move
the spec forward for
... Candidate Recommendation. I'm pretty new to this whole
thing. If there wasn't anyone working on it the WG would
... recharter without WebVTT but theoretically it could be
re-added the following year. For me the most important
thing
... to work on is an implementation report and get it done as
soon as possible.
Thierry: As you mentioned the
spec is currently in CR so the next goal is to exit CR and move
to Proposed Recommendation.
... You're probably aware of the exit criteria in the CR SoTD
section.
... We need two tests per feature so we usually demonstrate
that through an implementation report. I saw in your
... slide that Silvia sent the URL for that the report you have
is very similar to an implementation report. It would be
... good enough but from what I've seen it is just a subset of
the features.
... Therefore what the group would be looking for is a full
implementation report running the test suite against the
... four browsers. That would certainly help and having the
current status of the implementation experience.
... Also Silvia mentioned non-browsers and I would like to
understand if there are other products to use, such as
VLC,
... set top boxes or polyfills.
Gary: I think VLC is probably the
biggest one. JB told me at FOMS he had implemented the spec
fully. We would want
... to ratify that. If we can use VLC then that's one
implementation. Safari tech preview has most of the spec
complete
... as well and they seem to be onboard with updating things if
there are any bugs.
... Also dash.js and shaka and the vtt.js polyfill implement it
and we can update the polyfill as necessary.
Thierry: So the tests can be run on VLC?
Gary: Probably they would need to be run manually.
Thierry: OK. Just to clarify, we
are not looking for a product that implements the full spec.
You could have one that does
... half, another three fourths. We are looking for each
feature to have two implementations. A full
implementation
... is great but not mandatory, to clarify.
... Would you be able to provide such a report, running the
tests on the players?
Gary: Yes, the slides were just
the web platform test results. The rendering tests require
manual review but the
... parsing and API test results are complete.
... I can go through that and collate a report.
Thierry: That would be great and
would definitely help us to know the status of WebVTT against
implementation experience.
... If we are close to our goal of the exit criteria that would
be good news.
... Just to explain David has not much time so we are kind of
missing a Chair for WebVTT but I am ready to help. First
... let's start with this implementation report and understand
more deeply and schedule your attendance to calls as
needed.
... You can attend all the calls, it's not a problem, but we're
not going to talk about WebVTT every week I guess. We can
... schedule calls with parts dedicated to WebVTT.
Nigel: +1
Pierre: Gary, thank you for
joining today, it's great to see progress there. I think we
really need a Chair somehow. Gary
... that could be you. Don't quote me on it but it should not
be much work. I think we really need a Chair for this
effort.
... Thierry I don't know how you want to do this.
Thierry: W3 Staff are not allowed to be Chairs. If Gary wants to do that I can request that of course.
Gary: I would be willing to do that.
<inserted> Thierry: As Pierre said I don't think it's much extra work except for transition but I will help
Thierry: you do that as I do with Nigel.
Nigel: I'd be happy to schedule a call Gary with you and Thierry if you want to discuss Chairing more.
Gary: That's a good idea.
Thierry: Just one more issue is:
Gary are you aware of more comments that came in that we need
to address with tweaks
... in the spec?
... For example anything substantive that would need a new
CR?
Gary: I am not sure if there's
anything new. There might be something that is currently an
issue against WebVTT but
... I'm not positive if changes are needed or not. I can talk
with Silvia and find out.
Thierry: Okay. The last thing is
if we don't fulfil the exit criteria, let's say there are 2-3
features not implemented, they
... can be marked as at risk in a CR edition and removed to
allow us to get to PR. The Rec would then miss those
features.
... Those are also possibilities if we don't meet 100% of
features implemented.
Gary: That's good to know.
Thierry: I know that Nigel is
working Chairing the AD CG. I saw recently that it was put into
the Strategy Funnel and
... therefore I'd like to understand more about how you see
this work entering the TTWG?
... Would you like it in the next Charter?
Nigel: I think I've made this
point before - there's issue #4 on tt-reqs for example.
... The current status is there's a very rough draft spec for
the AD profile and there's one implementation that I'm
aware
... of so far, which is the BBC's.
... The AD CG has I think 19 members currently, one more joined
today or yesterday.
... My goal is to get to a point of "good enough" spec material
to bring it into TTWG in the May 2019 Charter. Since it
is
... a relatively small profile of TTML2, I am hoping we can get
to Rec very quickly.
... It's challenging but feasible I think to have the ADPT
profile at Rec by this time next year.
... Does that help?
Thierry: It sure does!
... Are there a lot of people in the AD CG who are not yet
participants of the TTWG? If that is the case would you
think
... that some people would join in the TTWG and bring in new
blood?
Nigel: Yes, possibly. I would
hope so, if only because from the current active membership of
TTWG I don't think there has
... been a lot of input. The one exception is the Editor is
currently John Birch who certainly was and may still be
an
... Invited Expert member of TTWG.
Thierry: That triggers a last
question: how do you envision the way this spec will be
specified?
... Would you like to work like we currently do with the Media
Track CG which does all the specification and editing and
... then bring it to the TTWG or would you like the CG to join
the TTWG and work within it?
Nigel: I would be open to
different models but my preference is there is one clear home
for this spec work and when we
... move into TTWG then TTWG is that home. It would still be
encouraged that AD CG members provide review feedback
... and of course they can provide implementations too -
there's nothing that says implementations need to be
provided
... only by WG members.
Thierry: That's a good point that
it needs to be clear where it's defined. I think it is better
to have it in the TTWG if
... the current participants are okay with that and if there
are new participants who come in and help that would be
great.
Nigel: Thank you everyone for a productive meeting as always, and apologies again for the late start. [adjourns meeting]