Meeting minutes
This meeting
Nigel: Today I only put one item on the agenda! Which is what we aim to do this year
… Any other business?
group: [no other business]
2021 Workplan
Nigel: End of last year, I asked what members want to achieve.
… Our group has fewer participants than before so I think it makes sense to try to focus
… on a small number of achievable goals (e.g. 1), do those and then move on to the next thing.
… Happy to have an open conversation about this.
… For me, coming off the back of the last meeting, I think there are several useful achievable things,
… and prioritising them is important.
… For example:
… * User customisation semantics for captions and subtitles
… * Just getting TTML2 to Rec?
… * ADPT to CR
… Those are just three from me. Can I open the floor - any comments or suggestions, or desires?
Gary: For me, it would be to move WebVTT forward.
Nigel: What do you mean by that?
Gary: There are a bunch of pull requests that need to be addressed.
… Basically get WebVTT into a place where we can move it towards PR,
… particularly with the live standard addition, with all the features pared down to what
… is deployed, so then we can start adding the features that should be in but
… are not yet available, after the fact.
Nigel: So marking some features as "candidate" status?
Gary: Yes
Nigel: On the user customisation front, I've implemented a prototype of all the options in CTA-CEB35 on top of imsc.js
… and it made me wonder if we need to define semantics for user customisation that can work interoperably across players.
Gary: For TTML and WebVTT?
Nigel: Yes
Gary: Having a single place that defines what possible user customisations should be would be great
… and we may want it to go into detail about how it applies to TTML and WebVTT as well.
… Alternatively we could just list out the customisation options and then WebVTT and TTML can reference that and
… say how they apply to themselves.
Andreas: Regarding the first, to gather information about how subtitle customisation should look, are you thinking of a general document
… that would be applicable to any captioning format?
Gary: Yes I don't see why not
… If it makes sense. We could always say "if it only makes sense to format X" for some reason, then mark it as such.
… For most of the standard stuff we're used to it doesn't matter what caption format it is coming from.
Nigel: One question is whether the customisation should be only at the presentation side or should be
… constrained to style attributes and properties that are already available in the format.
Andreas: This could be an interesting part, how to deal with customisation requests that aren't supported in the incoming format.
… And how to deal with customisation requests that would contradict the author's intention.
… Those are all cases that we meet in operation. If we see for example the existing options available in
… our OSes or TVs they may possibly not really look at the author's intention or what the incoming format is capable of.
Nigel: My go-to example for that is text colours, where a one colour outcome is too simple, and users in the UK anyway need to see different
… colours, and authors set colour as an intent. So I implemented a colour map customisation to set a palette.
Andreas: And there's a need for the user to say whether they want to override what's set, or only set if something is missing.
Gary: That's what the Apple model is.
Nigel: I find it hard to understand given that in TTML anyway almost all the style attributes have a default value, so how would
… the system know if the author's intent was captured by deliberate omission or by specifying something?
Andreas: We should start with concrete examples and match against common practices in different countries, and also try it out against
… existing customisation interfaces.
… From there we could get a feeling what direction we could head with our documents and try to work out some
… first recommendations.
Nigel: What I had hoped might be a small achievable goal seems to have escalated very quickly into a very big complicated
… goal that isn't easily achievable, which is probably reasonable!
Gary: You'd start off with a set of knowns, like FCC, CVAA etc and consolidate into one document, and work out what that means,
… rather than doing everything at once.
… Then get deeper after that.
Nigel: Yes, I think that's right.
Gary: Otherwise we'll never get anything done.
Nigel: CTA already did a lot of this, but at a lightweight level from a technical perspective, and there are also the MAURs.
… The question is whether this group should do anything beyond that.
Andreas: We should look at non-US sources too, like in Europe. In Germany all HbbTV subtitles are customisable but it's not mandated or written down.
Nigel: Something is written down though?
Andreas: Yes, for example HbbTV subtitles are based on user tests made in different EU funded projects.
… So there are sources for that.
Gary: Definitely this shouldn't just be a US-UK subtitle user customisation options.
Nigel: +1
Pierre: The first step might be to collect all this information in one place, if it is already done.
Nigel: That could be published as a WG Note very easily.
Pierre: From an implementation perspective the pitfall to avoid is doing something from one territory, and then
… someone says "what about here" and then the API gets more and more complicated.
… To deal with all the different models.
… If we could rationalise all the existing practices that would be ideal.
… Maybe start with a subset of all those practices. But in my mind the goal is to avoid
… 3 or 4 APIs, one for each style of customisation, if it can be avoided.
Nigel: I also wondered if there are any topics where we should have a "virtual f2f" meeting, maybe a couple of 3 hour calls in a week.
Pierre: If we can frame the customisation question in a good way then it could make sense to share that more widely
… and invite a wider group of interested people in to discuss in a "workshop" (but don't call it that).
Nigel: Agree
Nigel: The in-flight Rec activities that are not at "stage 1" are TTML2 CR and WebVTT CR
… Speeding up TTML2 would need us to go and do a load of implementation work, possibly.
… Gary has already explained the idea for WebVTT which sounds achievable.
… Are there any other candidates for things to work on?
Gary: There's working with CSS WG around viewport units in the video element.
Nigel: Yes, thank you for the reminder.
… The other CSS related thing is trying to advance e.g. fillLineGap, where it's in need of tests and implementation.
… I'm not comfortable saying that's the activity of this group, but it could be that the members of this group are
… motivated to push it on, and if that takes up their time it could be a bit distracting, albeit useful.
… From my own point of view, I'm struggling to prioritise subtitle customisation vs the AD profile of TTML2;
… in the BBC anyway it may be easier to get momentum from others on subtitle customisation, which may steer me.
Nigel: What do you think of the general idea of trying to narrow down our activities to a small number of deliverables and focus on them?
… Does it seem like a sensible approach?
Pierre: Can you summarise what's left in TTML2 to get to publication?
Nigel: It's completing the CR Exit Criteria for the Implementation Report.
Pierre: What's missing? Is it easy or hard?
… It'd be nice to finish that spec. These other things are interesting but they could be a lot of work.
Nigel: The CR Exit Criteria say no feature is added/removed so none is at risk, but 2 independent implementations are needed for each
… modified feature.
… The Implementation Report shows 1 passing implementation for all the validation tests, but only 1 for one of the presentation tests.
… If we had 2 implementations for every test it would be totally straightforward, but we're a long way from that.
Pierre: Is it possible to switch to the candidate feature mode? It's not good to have this hanging.
Nigel: I'm not sure we're allowed actually, to introduce that to an existing Rec.
Pierre: We could go back to WD and round the loop?
Nigel: Yes but we'd have to call it something other than TTML2, I suspect.
Pierre: We don't have to have two passes for every test?
Nigel: No, if we could refactor from features to tests we could possibly show that 1 validation test and 1 presentation test count as 2 passes.
… But right now that could only possibly affect one feature.
Pierre: I'll look at the presentation tests, it may be that we have some options.
Nigel: It seems to me from this conversation that we have quite strong motivation to push WebVTT and TTML2 on, and
… also good energy behind starting small with customisation in a way that can be extended later.
… And user customisation could be a good topic for an extended meeting of some form (extended in time and attendees).
Meeting close
Nigel: Thanks, this has been useful. Our next call is in 2 weeks. In that time I may be able to look at the TTML2 IR and pull it apart to
… work out what we really need to achieve to meet the exit criteria.
… Thanks everyone, happy new year again, see you in 2 weeks. [adjourns meeting]
group: [general wishes of happy new year]