Timed Text Working Group Teleconference

01 Feb 2019


See also: IRC log


Andreas, Pierre, Thierry, Glenn, Matt, Marco_Slik, Frans, Gary


<scribe> scribe: nigel


Andreas: Welcome everyone to this joint meeting of the W3C Timed Text Working Group and EBU Timed
... Text groups.
... I chair the EBU group, and Nigel is a chair of both of the groups.

Nigel: Do we all know each other?

Andreas: We mostly met yesterday but introductions could be useful for some people.

Marco: I'm Marco Slik, an R&D engineer at NPO in the Netherlands involved in the very broad spectrum
... of playout distribution and access services, I joined the EBU group a few years ago.

Gary: Hi, I work at Brightcove, we do online video with video.js and the web player and I focus on WebVTT

Thierry: Thierry Michel, W3C team contact for this WG

Nigel: Just an admin statement for me, which is that as a W3C WG meeting any IPR contributed here
... needs to be cleared. TTWG members have already made this commitment, if any non-TTWG members
... contribute any substantive material that ends up in a W3C Recommendation track document then as
... Chair I will need to get a similar commitment from the contributor.
... That's the formality over, in terms of IPR!
... Can I check everyone is aware of IRC and the minuting of this meeting?

Andreas: Yes, I will send the links around too.

Nigel: Just to let everyone know that this meeting will be scribed on IRC and that the log will be
... turned into public-visible minutes after the meeting. If you do not wish something to be minuted
... please tell me before you say it!

This meeting

Andreas: One of the reasons we meet together today is the shared interest in live TTML.
... We had a discussion at the TPAC meeting in Lyon a few months ago and also some other discussions
... in the EBU group about this.
... There was a discussion of the submission of EBU-TT Live to W3C. At TPAC we decided we needed
... more information about the plans of the groups regarding live and standards activities.
... So we thought this meeting would be a perfect opportunity to discuss this.
... The main part will be dedicated to this topic.
... By introduction we have some example from operation, Nigel from BBC, and the open source
... activity of the EBU-TT Live Interoperability Toolkit.
... Matt from Red Bee, and Marco from NPO and I will bring in requirements from some German
... broadcasters.
... After that we will look at the live standards activity.
... Also as we're here today we can widen the scope a little bit.
... I talked to people before. My interest would be to check how much we can bring TTWG and EBU TT
... activities together and doing more work in W3C.
... [missed a bit]
... We have concrete examples of multiRowAlign and linePadding.
... The order is not fixed yet, we have a list of topics on the wiki page.
... We can discuss where to start.

Nigel: Thanks for that, as you have said we have a list of topics on the wiki page.
... Looking at the schedule, what should we look at first?

Andreas: Let's begin concrete with EBU extensions and then dive into live.
... Then come back to cooperation and collaboration of the groups at the end of the meeting.

EBU Extensions to TTML

Andreas: Some information I shared

-> [pointer to Andreas summary mail] Andreas's summary text

Andreas: The main question is if the two vocabulary items ebutts:multiRowAlign and ebutts:linePadding
... If it makes standards work of W3C easier we would be able to allow them to be maintained in the TTWG.
... If there's interest in W3C group to do this.
... Otherwise we can keep it in the EBU TT group scope.
... I sent an email to the EBU reflector making a proposal, and have had no objections to that proposal.
... From whatever date, multiRowAlign and linePadding are maintained and at the end controlled no longer
... by EBU but by W3C. EBU will not make any changes to them.
... If there is another EBU spec then it can reference any updated definition in a W3C spec.
... This allows easier maintenance and adoption of bug fixes found during W3C activities.
... The general idea is to update the items for clarification and bug fixing.
... If completely new requirements come up that are not satisfied by them then it may be better to add
... two completely new items in the TTML namespace. If for some reasons semantics change then it is in
... the interests of the EBU group that they are backward compatible with the existing EBU definitions.
... If W3C has control then they maintain it and it is their decision process how to deal with updates.
... We currently have at least 4 members who are in both groups then I do not see any conflict.

Matt: Is there any precedence for this?

Frans: I'm not aware of this exact case.

Glenn: When you say hand over control are you assuming a namespace change?

Andreas: No

Glenn: The proposal is a syntactic transfer but cede control to the W3C of that subset of the EBU owned
... namespace?

Andreas: Yes
... Pierre as editor of IMSC, the question is if there is any benefit. I think it is clear it would be maintained
... in IMSC scope.

Pierre: There's always a benefit in having a single forum to address issues.

Glenn: If the formal definition moves to IMSC, would it be feasible to define, say in TTML3, two new
... attributes of the same name but in the TTML namespace?

Pierre: To me that's a separate issue.

Glenn: Well they're both in W3C control

Pierre: That's true today

Matt: Would they be equivalents?

Glenn: I've objected to including foreign namespaces in TTML in the past.
... That's not true for IMSC which is a profile that adds vocabulary from other places like EBU-TT.
... That was completely appropriate in IMSC, but it has never been the case that TTML refers to foreign
... namespaces. If we do this then we should have a plan or intent to bring them into the TTML style
... namespace and define them equivalently.

Pierre: That's orthogonal, we could do that today.

Glenn: If you think that's possible, then that's okay.

Andreas: Your proposal is to duplicate semantics but give it a different name or namespace?

Glenn: 1. Move formal definition into IMSC, same namespace, words etc so it is consistently defined.
... 2. Define the same names and semantics in a TTML namespace. At that point the IMSC spec could be
... updated to refer to the TTML namespace semantics if it wants to and also say it is available in the EBU
... namespace.

Andreas: This has been discussed in the past and EBU has said that is not the intent to duplicate existing
... vocabulary items, and that duplication makes no sense. Personally I think that is the best strategy.
... We use the EBU vocabulary items because they are established, implemented and we don't want to
... make other implementations have to change to a new vocabulary with no need.
... This is formally not needed but it goes together. If the overall intent is to duplicate the EBU namespace
... items without adding new functionality that would not really justify bringing over the control of the
... EBU namespace items. We had this discussion in the past and the position has not changed.

Glenn: We have to recognise that IMSC and TTML are different and have different conventions.
... We have followed different restrictions. As a primary language definition it does not reference any
... foreign namespaces, just the core XML namespaces and those defined in TTML itself.
... Other than this potentially we have never discussed a proposal or accepted one to refer to foreign namespace.

Pierre: I don't see why we need it.
... W3C could decide today to create a duplicate attribute. We've had that discussion. I don't see what is different.

Glenn: Bringing them in to IMSC will not satisfy requirements for TTML that are not IMSC users.
... Everything is in the TT namespaces.

Pierre: We can have that discussion (again).

Glenn: I have no objection to making use of the EBU technology. It is great, proven, deployed, and I don't
... want to seem like I'm denigrating EBU in this regard.
... As Editor of the primary core language definition I have to stand by certain principles of language
... consistency and this is one I'm going to insist on.

Andreas: Possibly it is best to follow Pierre's proposal and concentrate on what is on the table.
... This would be to hand over control of the items multiRowAlign and linePadding only into W3C for
... adoption by IMSC and nothing else.
... What is the opinion of the other members present?

Matt: I think it seems sensible. The one question for me is the overall strategy and context?

Frans: Pragmatism

Matt: It doesn't make sense to have multiple definitions of the same thing and then the community would
... need to do the same work twice.
... We would like to contribute to the biggest circle in the venn diagram unless there's a proposal that is
... hyper-specific. We would like to break down the notion that captions and subtitles are different from
... country to country because in my experience they are not. Interoperability is the goal.

Frans: I totally agree, this makes sense and we should be pragmatic not thinking about not invented here
... concerns.

Andreas: I see nodding from Marco, no objection from W3C side.
... I will share the proposal from the EBU with the Timed Text group.

Pierre: I think a liaison is needed, and an issue on IMSC to adopt in the next version.

Nigel: I'm not sure if a liaison is needed.

Pierre: For future generations a standalone piece of paper that explains what happened would be useful.

Andreas: I can make a proposal offline and check it with both groups to go forward.

Nigel: That is okay for me too.

Overview of EBU standards

Andreas: [shows list of EBU specs]
... EBU-TT Part 1 v1.2, the base of every standard we published, intended originally for exchange and archive,
... as the successor to STL.
... Not in TTML2: linePadding, multiRowAlign
... Not in IMSC: smpte, clock timebase, markerMode, clockMode, cell length metric except for linePadding
... EBu-TT-D, same scope as IMSC, a subset of IMSC. Includes multiRowAlign and linePadding.
... EBU-TT Live, subtitle contribution over IP. As well as multiRowAlign and linePadding, there are four
... attributes not in TTML2 or in IMSC, sequenceIdentifier, sequenceNumber, authorsGroupIdentifier,
... authorsGroupControlToken.
... EBU-TT Part M is definition of metadata elements, no compatibility issue because TTML and IMSC allow
... adding metadata.
... Also we have some complementary documents.
... A recommendation how to map EBU STL to EBU-TT Part 1.
... A recommendation for carrying EBU-TT-D in ISO BMFF.
... A carriage spec for carrying EBU-TT Live over WebSocket.
... A question I have is for not just EBU-TT Live, but also part 1 and EBU-TT-D how they could converge
... with the TTML specs. Later we have an EBU group meeting, this will be one part of the discussion there.
... Looking at the minimal part that is not inside IMSC or TTML I see some option really to possibly
... make use of some W3C specs to satisfy the use case we originally thought of with part 1 and Live.
... EBU-TT-D is already solved by IMSC, EBU-TT-D is just more restricted.

Glenn: The idea is to identify areas of difference as you have summarised, and then find ways to...
... I guess I'm not clear on the next step - update the EBU specs? Or some action on W3C side?

Andreas: The background is the situation that if you go to the industry and ask about information about
... TTML they say there are so many profiles they don't know which ones to support. They don't implement
... a complete set. That's not a good situation, so I think it would help to limit the set of standards that
... needs to be implemented, and I really like simple solutions so in my simplified view, for example for
... part 1 we could say for production and exchange, just use IMSC, but then check if everything in IMSC
... now satisfies this use case, and if not, make an addition or some other change so that, out there, where
... we need implementation, people know they just need to implement IMSC.

Glenn: Thinking aloud, I'm wondering if part of the problem with all these profiles is that very few clients
... if any actually do the profile processing that is defined by TTML in the sense that it actually verifies
... if the profile asks for features to be supported and acts accordingly if they are not.
... I felt always bad about the history of the EBU's adoption of TTML when it ruled out use of the profile
... system which I thought then and still think is a mistake because it removes the ability of the client to
... refer to the author defined profile information. I'm not suggesting anything, but it may be that the fact
... that very few clients use the profile is part of the problem.

Andreas: Try to look forward not backward.

Glenn: I agree.

Nigel: Chicken and egg problem - if profile adoption is not there then it is hard to specify it, and TTML
... implementation is already big enough without including profiles.

Andreas: Question for Pierre, do you see a path to adding EBU-TT Part 1 syntax into IMSC if there's a gap?

Pierre: We can look at it, it is not worth creating another flavour/layer/profile but if they are useful
... in production only but not distribution then maybe there is genuinely a need for another flavour.

Frans: There has been little drive to change the production side, more the distribution side.
... So EBU-TT has not been well adopted.

Matt: We've made a lot of EBU-TT output, everything for BBC since 2014. We probably could derive some
... useful stats on the number of EBU-TT documents.
... Ofcom has recently published a statement about the extension of accessibility regulation to online
... video platforms, which included something about standards support. There was an attempt to try to
... persuade the regulator to add guidance or stronger to direct content providers to use particular formats.
... We have one client for whom we make 13 or 14 different variants of the same data. That's driven by
... probably no good reason, but the proliferation has caused fogginess. Anything we can do to reduce
... that will be very valid and valuable. If you read that statement, [missed a bit] those are not small
... amounts of money, having one format that could be propagated to every platform would be very
... useful.

Andreas: Useful comments. From my experience IMSC for example [missed a bit]
... but the value of having one format overweights possibly the extra work for features that may not be
... needed. Most of the things we leave out are not really a problem in the implementation. It is more the
... parts we brought in, with multiRowAlign, lineGap etc are the tricky parts.
... To actively give the message that you need to implement full IMSC coming from different standards
... organisations would definitely help I think. One of the main questions we can look at already for
... IMSC is timecode. We had discussions with lots of people, lots of stakeholders are not yet ready to
... go away from timecode.

Pierre: And a lot of them use it badly. We really have to understand the use cases extremely well.
... In the modern world timecode will not work 99.99% of the time with component based media.

Frans: I can only agree it must be use case driven.
... That's the most important for production.

Marco: When you want to transition you have to support some things... [missed a bit]

Pierre: It will not work with online media.

Marco: I will not manage to get away from smpte timecode in 3 years, but we could maybe start the timecode
... clock at zero for example.

Pierre: I don't think saying timecode can not or must be used is the answer, it's a lot more nuanced.

Andreas: I do not think we can say we see a better future and will not help you now.
... For German broadcasters we need an incentive to change from STL.
... We thought we did that with EBU-TT but they did not use it.

Frans: When we started the work the premise was that STL can not do the things we want but in
... production people have not hit that wall yet.

Matt: Yes, we have not hit that inflection point. We're getting more foreign languages, like Mandarin.
... We can't do live captioning in Mandarin, in 608 or Teletext, but we need an answer before people
... invent something that won't work.

Glenn: A character coding issue?

Matt: Yes but the broadcast standards don't support those characters.

Pierre: Contribution of HD single language content is a solved problem. We're not competing with ATSC or 608.
... We're talking about worldwide online distribution of content.

Matt: Precisely. We're working with an online entity that has no SDI platform.

Andreas: That's a good start into the discussion, I propose we move over.

Pierre: On that topic, since we're all together, if we are serious about trying to rationalise this it would
... probably be a good idea to set aside a day at a workshop to really get to the bottom of it, and get
... people to tell us what they are doing and want to do otherwise we will make a wrong decision.

Frans: We tried to do this a number of years ago for EBU-TT Live.

Pierre: For instance EBU-TT Part 1, who are the target users?

Andreas: Everyone who uses STL.

Matt: I can collate data from most UK broadcasters.

Frans: I do think there is point to verifying proposed pain points with users.

Andreas: From the experience of the live workshop it was a bit chaotic and was hard to bring together.
... It was good at that time, but as you say everyone goes to their domain and community and we come
... back maybe in the same kind of circle.

Pierre: Next opportunity is maybe IBC.
... It would be awesome to end up with a single format across the entire chain, not for old applications.

Andreas: yes and if we are committed I'm sure we will get others onboard.

Matt: Clients ask us how to give flesh to the ecosystem

Frans: We need to give a roadmap

Matt: And state the benefits.

Nigel: That Live workshop was in the same room as this meeting in 2012, and one of the reason it was
... chaotic was that people wanted to address not just the workflow we had in scope but at least one step
... further, like speech to text engines.

Frans: I agree, we need to focus on the use cases.

Live subtitle contribution

nigel: [describes EBU-TT Live and the Live Interoperability Toolkit] - Frans took notes.

<inserted> BBC subtiltes 100%, includes all live programmes.

<inserted> Stateful Teletext based (legacy) technology is used for this.

<inserted> Does NOT support future requirements, such as live streaming of bespoke events, cloud encoding, etc.

<inserted> When studying the standards supporitng this authoring-to-encoding part of the workflow.

<inserted> EBU-TT was the obvious choice (the BBC was already producing in EBU-TT).

<inserted> EBU-TT-Live (Tech 3370) specifies how you can send a stream of text chunks to facilitate this scenario; with 0 or 1 documents 'active' at any time.

<inserted> The EBU-TT-Live Toolkit validated the spec and helps implement it: https://github.com/ebu/ebu-tt-live-toolkit

<inserted> Timed changes do not have to be explicit ('just send a new document') as was demoed in the SwissTXT demonstration at EBU PTS 2019 for example.

<inserted> SMPTE timecode is not allowed (as you cannot do match on it).

<inserted> Carriage using web sockets has been defined.

<inserted> EBU-TT-Live / TTML carriage in RTP for use in SMPTE 2110 ("IP replacement of SDI") is being worked on; BBC has submitted an RFC for TTML carriage in RTP.

Matt: Background to fit into that.
... [shows slides]
... We're a subtitle vendor, and a few years ago wanted to replace our live captioning platform. We were
... surprised at the lack of forwards thinking in the market. I was in this group already at that point and
... could see that the world would change and Teletext would not work.
... We built our own java based captioning tool called Subito.
... Effectively internally it makes TTML.
... Can output Teletext, Nufor, 608, Cavena, EBU-TT Live, and can do those simultaneously, applying
... the constraints in real time.
... We did our first presentation at PTS 2 years ago, and did some work using the LIT. We had a basic but
... brittle proof of concept. More recently using for our customers.
... Development challenges, managing pre-conceptions of HTML and XML ignoring good EBU documentation
... ignoring format details because of lack of understanding of the use case, so deviating from the spec.
... Partly caused by lack of reference examples and material.
... Bad habits like using line breaks for positioning.
... Ignoring test materials that are free like the Interop Toolkit.
... Challenges about reformatting, we focused on conveying the intent of the captioner to the stream.
... But how to cope if presentation format is a different aspect or applies other constraints?
... In the implementation side we didn't think enough about that. No practical constraints, but need to
... avoid developers making inappropriate solutions.
... We created a multichannel Distribution Node. Originally we required consumers to connect in, what we now
... have is scalable and supports multiple destinations.
... Testing is difficult, need a 24/7 delivery destination.
... Security: not an issue with 608 or Nufor, wide open or limited in format.
... We've implemented Auth0 with security based on IPv4 lockdown approach.
... Haven't implemented handover functions, the idea that you may have more than one source of
... caption documents, and we have a model in Part 3 for passing over control seamlessly so everyone
... knows who is in control and where their data is going. Needs more ecosystem, lacking at the moment
... for testing and demonstration.
... "flying hours" - if we have used libraries with memory leaks etc then eventually things will go wrong.
... Until we have solid real world use we can't be 100% sure it is robust.
... Used for Department of Parliamentary Services, Australia with a web feed.
... Extended proof of concept, due to launch imminently.
... Second is "Intersub", real time transcript service. Currently Newfor, plan to migrate to Part 3, which
... will simplify the stack and remove newfor handling deduplication code.

Andreas: Really good, thank you, very interesting!
... Next up, Marco.

Marco: [trouble presenting]

Andreas: I will take over and we come back to Marco.
... Our broadcasters are fine most of the time. They look for new solutions if current tech cannot solve
... their use cases.
... One problem in the past is live streaming of broadcast programmes with closed captions, not burned in,
... on the distribution channel.
... All of the chain is standardised especially at the end. There are stable and good standards to package
... EBU-TT-D and IMSC into MP4 and deliver it with DASH and HLS.
... To encourage implementation we did a demo that has been online for 1 or 2 years where we use the
... live broadcast signal and take out the subtitles and bring it into IMSC and deliver it over HLS and DASH.
... [shows live Bavarian broadcaster BR's output]
... Teletext subtitles translated into IMSC.
... How it works: for the encoding and packaging, receive SDI with embedded teletext subtitles in the VANC.
... They decode it and then translate it to TTML and package it in MP4 for the client. That works.
... But what if you have a non-broadcast workflow, for an online only platform with no broadcast chain.
... And our broadcasters say they have a couple of internet-only live events and they do not want to buy
... whole broadcast systems for these events to subtitle them. One way it could be done is that at the
... packaging side you receive the video and audio together and synchronise it and deliver as MP4 to the client.
... The scenario we would like to solve is to use EBU-TT Live for the contribution to the encoding layer.
... Also we would like to get people together like we did in the past with this CMAF/DASH demo.
... The challenge we have at the moment is to find a packager and encoder that would implement
... EBU-TT Live and the synchronising mechanism.
... It would also be possible to use a legacy format like Newfor and transmit teletext information.

Matt: Newfor is not an open standard. Everyone who has a copy of it has inherited it rather than been
... able to obtain it!

Andreas: And all of them use a different variation of this protocol.

Pierre: Who created Newfor?

Matt: Sysmedia, a guy called Andrew Lambourne who now works at the University of Sheffield Hallam I think.
... It's a simple serial style wrapper around the core Teletext page instruction, with a session, flow management,
... and wraps up what literally should be on the screen.

Andreas: What we see as a chance is the next Olympics. We have heard they want to offer much more
... online than on broadcast, and they need to provide captions. That's the goal and we are really looking
... for a good working solution here.

Matt: We did some work with Channel 4 in the UK, who broadcast the paralympics. They wanted to make
... it all accessible online and the only solution was to hire a crate of kit from the US, go to SDI and back
... out again, and doing that all cost quite a lot of money. The requirements really do exist.

Andreas: Marco, can we hand over to you?

Marco: [presents slides]
... [Status subtitle market]
... Global overview of what you all said, to make clear the Europe subtitle volume is extensive.
... Most countries have regulation with additional European regulation that will probably end up with
... having 100% hearing impaired subtitling. About 50% of Europe uses translation subtitles as well.

Pierre: Checking my assumption, 5 years ago a lot of subtitling for translation but there is no [missed]

Frans: Different per country

Marco: The view is that accessibility needs to be at a higher level.

Frans: Increasing pressure, depends how far people can go because of money.

Pierre: On the accessibility standpoint, what is the coverage in Europe?

Frans: Subtitling only (captioning), 100% in NW Europe.

Marco: 95% on all content.

Matt: Many countries it is only the public service broadcaster.

Andreas: In Germany main channels 100%, major public ones, commercial ones down to 20%.

Marco: Will change with new EU regs.

Frans: Not sure timing

Matt: Equivalent of the US regs

Pierre: Public and commercial broadcasters?

Frans: Not equal

Andreas: Media industry requirements not so strict.

Matt: Strange things - some countries have more stringent regs because of their governments but
... tech affects ability to deliver.

Pierre: Statement that penetration will increase is because of regulation or cost?

Marco: Regulation mainly.

group: [everyone wants to see data on that!]

Marco: On the translating part, a lot of countries have burned in subtitles too.
... Differs between countries, in Netherlands it is all foreign language.
... [Subtitle market status]
... Shift from linear broadcast to on-demand and OTT.
... New requirement for different devices not knowing Teletext.
... Teletext used in broadcast facilities, page 888 for the viewer, being phased out.
... Equipment is end of life, end of support, no new companies offering solutions.
... Typically something I run into, no Teletext support in encoders.
... Current support only goes to 7 lines, I need 15 or 16!
... Current tech support is cut off but IP protocols not there - we really need to do something new.
... New players on the market, mainly internet startups.
... Know a lot about new world interfaces.
... Need standardisation to get this streamlined.
... Increasing demand for good access services and customisation on client side, so not sending
... a rendered subtitle image but sending text so the front ends can customise.
... New volumes - non-audio videos without subtitles
... Looking at new functionality for web players, extra metadata etc.
... [EBU-TT as replacement]
... File based, not for today. Teletext based - all proprietary, being phased out.
... [Live transport in the chain] - picture from EBU spec, slightly modified.
... Live part from authoring facility to playout, then playout to encoding.
... EBU-TT Live -> EBU-TT-D/IMSC for player rendering and DVB
... [NPO's live subtitle chain]
... Live subtitle department done with one editor for simple programmes, and 3 editors for more complex
... talking heads programmes, with respeaking, correcting and cueing, with video delays. Very demanding
... for respeaking, so change editor roles during the programme.
... Also trying to do automated subtitling, really a challenge at the moment.
... Then playout, file based with STL and Cavena at the moment, need to go real time to encoder.
... As Andreas said, Newfor and Teletext are the old world. Looking toward having a new world approach
... where styling and formatting are equal and extendable through all devices.
... So we want block based subtitling, handover between editors, time relationship is SMPTE based but
... not always 'do it now' but could be late file delivery with overlaid live subtitles, then ingested during
... playout for replay later.
... Looking at media timecode in next generation which should last at least 3 years.
... Need to fan out to multiple destinations.
... Timebase processing (delay compensation)
... Lowest delay NOW (remote capability) is needed
... Same for file based playout.
... SMPTE timecode, timebase correction, fan out to multiple destinations, timed metadata.
... A TTML based approach really important for timed metadata.
... In encoder, fan-in from multiple sources, then synchronise ABR and DVN stream transport,
... subtitles and timed user metadata,
... join other worlds like HbbTV.
... [Needed] Widest standardization as possible.
... Currently have momentum to get rid of old world possibly, but tools not available for new world yet.
... Really frustrating.
... Good to mention need for multiple languages including right to left. We have a huge amount of Arabic
... in Holland.
... Near real-time in parallel to SMPTE 2110 ideally in future, encapsulated in distribution for client renderers.
... I hope to have some support for aligning positioning especially for converting from Teletext to new world.
... EBU-TT Part 2 includes some guidance.
... Would like to do the translation as early as possible in the chain.

Andreas: Thank you for that honest but realistic view of what is happening and what the challenges are.
... Propose a break now for 15 minutes maximum.
... Return at 11:05.

Nigel: Thank you for those presentations - they were really great!
... [return from break]

Andreas: Can we start with a summary from yesterday?

Nigel: Not sure exactly what you're referring to?

Andreas: Yesterday it came up what could be done with EBU-TT Live and how to handle it in W3C.
... TTWG will have a new kind of layout a bit like CSS, where some functionality is taken out of the main
... specification and put into modules. This opens up a more flexible approach to address current
... requirements and parallelise activities so they do not block each other.
... In this context one of the proposals was to take the functionality and possibly the vocabulary of
... EBU-TT Live in a module published by W3C which then can be used by any profile of TTML, could be
... EBU-TT, IMSC or whatever.
... From my impression that's where it ended, it's a feasible approach.
... There are other questions how it could look at the end and the mechanics e.g. a submission from EBU
... as a W3C member. It is a feasible approach.

Glenn: Clarifying my thinking on modularisation approach and how it is not exactly like CSS.
... In CSS they carved up the existing functionality of CSS2 into pieces and called them modules.
... I'm not assuming and would oppose doing the same thing here with TTML3 because there are so many
... interlinked semantics between the different components and it would create quite a bit of effort to
... carve it into pieces and I don't see any advantage in evolution because it is fairly stable.
... What I see happening is in the context of TTML3 creating a new framework for modules to allow
... new functionality to exist that is not in TTML2.
... To the extent that we can put it into different module documents proceed on that basis.
... It may turn out that there are some substantive changes not in TTML2 2nd Ed that are appropriate
... for making in the core of TTML3 that are not in modules, but to the extent that we don't separate
... functionality then that would be the best approach, allowing parallelisation of work and editorial activities.
... Especially for new well identified groups of functionality such as live.
... Just communicating my understanding because that's what I'm going to operate on.
... If people differ from that I'd like to hear about it.

Andreas: I think it would work well for this use case, and it would also be good to discuss why we have
... the discussion. Right now EBU-TT Live is complete and is out there. What are the benefits of bringing
... into W3C?
... 1. To encourage vendor support, by having a global solution.
... 2. I know there are ideas and requests for updates of functionality. Here it is better to do it in one
... group rather than separate groups.

Nigel: Add one more, which is it would allow us to make a thing like IMSC Live, which currently seems to
... be excluded by EBU-TT Live even though in principle it is possible to make IMSC-conformant EBU-TT Live
... documents, I don't think that seems obvious to the wider world.
... Also there are some useful constructs permitted in IMSC that are excluded by EBU-TT Live.

Andreas: That's right, and you also point to an important issue that EBU-TT Live is first a description
... of how a live TTML workflow should look, with some additional syntax and semantics.
... It is bound to EBU-TT syntax definition, so limited to what is in EBU-TT Part 1.

Marco: Two main topics: The transport mechanism and the syntax of the documents.

Andreas: Right.

Nigel: In the EBU specs we've tackled them as separable things.

Andreas: So 3 things: Workflow, Carriage mechanisms, and the payload.

Marco: What I meant was the transport and how we encapsulate the handshake in mechanisms.

Andreas: There are clear benefits also in the EBU group, and everyone involved in standardising EBU-TT Live
... is convinced of the usefulness here.
... It would be useful to hear from others in TTWG, Pierre, Glenn.

Pierre: Yes, I don't have a very strong opinion about applying only to IMSC or to TTML as a module, I'm
... happy either way, whichever works best.
... I think it would be good to apply to IMSC somehow if it is possible.
... Last time I saw it it was merely a couple of additional attributes with no impact on visual presentation.
... The only feedback on the document that I have already shared is it is too long.
... Otherwise it seems like a good idea. I like the idea of trying to focus development on the technical
... specifications in a single place. The requirements can come from other groups.

group: [general agreement]

Frans: I totally agree, that is the main point.
... A second point is ease of marketing, branding, so considering a complete IMSC based future could be
... very attractive as the main thing people know about. We should embrace the best candidate we have
... and it is IMSC, that gives us the biggest chance of adoption.

Pierre: Specifically on EBU-TT Live and authoring, with SMPTE 2110 and the brave new world, it would be
... really good for this community to help folks pick the right technology, or someone will pick SMPTE-TT
... and say that's what we're doing now.
... My feedback to Nigel is the more we can do to narrow what goes over 2110 the better otherwise we
... will have disappointment, otherwise someone will implement TTML over 2110 and it won't work
... between vendors.

Matt: I would like to see removal of specific protocols for individual markets, in favour of global ones
... To avoid lossy transcode, reworking, and support new languages.
... I've had conversations about Punjabi, Hindi, Mandarin, Cantonese (bizarrely), Arabic, they all need some
... kind of kludge.

Pierre: Can we talk about 2110, how can we contribute positively?

Nigel: On the point about TTML vs IMSC for this, there are other uses of TTML than subtitles, which
... may benefit from live contribution, like AD, as agreed yesterday, so it makes sense to put it there.
... On the 2110 issue, have to use TTML because that is what has the IANA registration.
... Also want to encourage use of codecs parameter from the beginning to avoid the non-adoption problem.
... Also we will need other specs like in SMPTE, NMOS and EBU to complete the suite.

Frans: [scribe missed]

Andreas: We are here because we want to work more closely together. This work counts the same,
... need to focus on working together and make sure as Pierre says that the major stakeholders align
... on one approach to solving a specific problem.
... Parallel solutions won't help.
... I'm not so familiar with IETF. Two forms of collaboration. One is publish something and let the domain
... submit some feedback, send comments back. This works not that well in the past because of workload.
... The other is getting together in a room and talking through the problem.
... I'm not sure about this particular work, BBC is submitter so that is helpful. The relevant people
... should come together and at least discuss this paper and agree it is the right approach.
... For SMPTE I'm not sure.

Nigel: I've already received a request to discuss in W3C TTWG and am happy to do that.

Frans: Putting together a timeline would be useful.

Pierre: To better understand 2110, there's a bunch of specifications for audio and video but is there any
... about timed text today?

Nigel: No there isn't yet.

Pierre: I'd like to come back to that.

Andreas: Okay, from the feedback so far I haven't heard anybody objecting the approach to bring the
... EBU-TT Live activity into W3C at least to mirror the most important parts of EBU-TT Live as a module
... in a W3C spec.

Nigel: Is it clear to everyone what the syntax, semantics and general approach is?

Andreas: We may not have time to go through it.

Nigel: Just want to check if there is enough awareness to make a general decision.

Glenn: What's the decision?

Andreas: To bring the EBU-TT Live work into W3C

Glenn: Sounds good to me, mutatis mutandis.

Frans: To make IMSC Live?

Andreas: First to bring into TTML, second should we say it is our goal to make IMSC usable together with
... this module, if it is a profile or an addition to IMSC has to be found out.
... The other thing I think is also important is to bring over these two attributes like multiRowAlign.
... Do we say the EBU-TT Live attributes are now handled in W3C space?

Glenn: A more practical question - do we have a prospective editor?

Andreas: We clarified this yesterday Nigel will be.

Pierre: Do we need to make a decision today?

Andreas: Let us see if we are ready for a decision!

Pierre: On the namespace?

Andreas: What do you mean? Concrete proposal: move the syntax and semantics into a TTML3 module
... without changing syntax, then add to IMSC Live.

Glenn: I mentioned yesterday I would be amenable to enabling foreign namespace vocabulary in
... modules, so that could work for this and for multiRowAlign and linePadding.
... That would also in my mind be a potential incorporation of those into the TTML core but that is a
... future process to think about. By then it may be that we have acceptance of the use of foreign
... namespaces in TTML.

Pierre: On the relationship with IMSC, maybe the jury is still out, but my understanding is that these are
... timing attributes added to a TTML document?

Glenn: Almost like metadata.

Pierre: Exactly, they're permitted additions that don't affect presentation.

Andreas: Apart from timebase clock

Nigel: Like a pre-processor that affects the time during which the document is active, otherwise yes the same.

Pierre: We can just do it.

Andreas: Both a technical and a marketing side to this.

Frans: So we could end up in an ideal case in a few years with everything IMSC and the world is simple,
... that would be the best solution, forgetting TTML, EBU-TT all the rest of it.

Matt: It is being adopted already without a second thought, seemingly, so it is free of geographical boundaries
... already I would say.

Andreas: I don't think we need a final decision on this depth, but if we agree then we should prepare the
... next steps and propose it and do something similar as the two styling attributes.
... For me it is clearer we should handover maintenance and control of this to W3C.

Nigel: If we are writing a liaison, a piece of paper, we can either cover all of it, or have two, one for
... styling and one for live.
... I've heard enough that there is consensus on our requirements issue to take it forward for this year.

Pierre: A member submission would be the best thing.

Nigel: Quite a lot of overhead in doing that, I'd rather start in W3C with an empty thing.

Andreas: Proposing to hand over control over EBU-TT Live to TTWG, to republish EBU-TT Live as a module,
... further updates of EBU-TT Live done in the TTWG and not the EBU Timed Text group.

Nigel: EBU should have the freedom to iterate its specs if it wants to do so.

Frans: Query if we want to update EBU-TT Live or brand everything as IMSC Live.
... That's an important consideration. I would be in favour of steering everything towards IMSC.

Nigel: I don't disagree, maybe EBU should publish something explaining this position to the world.

Andreas: We are making an important shift, we may need to discuss this further. For me it makes more
... sense to make a cut and say we transfer it now to another organisation and will not update further.
... For the two vocabulary items we hand over control and if there are further EBU specfications we can
... publish a new spec referencing.

Frans: What if there is a bug in EBU-TT Live?

Andreas: If there is a short term need then we can do that otherwise we should not.

Frans: I am not sure what timelines we are talking about.

Marco: Isn't it about securing EBU requirements in a future spec?

Frans: I feel confident about that.

Andreas: We have enough shared members, we can collect requirements from the EBU and bring it into
... the new W3C spec but do not publish in EBU scope.

Marco: That makes it clearer to the rest of the world.

Frans: It is all about simplicity. In spirit we all agree.

Andreas: Long term goal is agreed, need to figure out how it would work in the short term.

Nigel: Checking in on the agenda, we've covered almost all the topics, 15 minutes to go.
... Pierre asked to come back to SMPTE 2110, and Andreas you wanted to come back to something too?

EBU and W3C collaboration

Andreas: [shares screen]
... We have these two attributes, and part 1, and EBU-TT-D and part M.
... Starting with EBU-TT-D, one approach we could take that would benefit industry is to say we stop
... work on EBU-TT-D and point to IMSC for any future update, or missing functionality or bug fixes.
... That would be a clear message.
... We should try to elaborate if we can do something similar for Part 1 and Live.

Nigel: I think a key point here is how we handle smpte and clock timebases, which are in EBU-TT and TTML
... for a reason, and not in IMSC also for good reasons. Can media time only work in playout scenarios?

Marco: As a broadcaster, I can say it won't work for the next 3 years.

Andreas; Stick with what works and is available.

Pierre: Exactly, that's what you have to do.
... You can adopt EBU-TT Part 1 today and constrain it. 3 works is a realistic horizon to work with
... vendors and service providers and say in 3 years we plan not to use timecode for those workflows.

Marco: We always have legacy from our archives so we need some timebase transcoding somewhere.
... It's already on my list. I think I'm forward looking within the range of broadcast views!

Frans: Absolutely. This is a very critical thing, we need to spell out the transition scenario, how does this
... work, what do you encounter in practice. If we cannot we are not in a good situation.

Marco: We have to guarantee that what is in EBU-TT now remains the same in large part for IMSC.

Pierre: Looking at the differences, is that the sum total?

Andreas: Those are the main things on a quick look. The really substantive things are timecode and cell.
... Cells are manageable by translation to %.

Pierre: Especially with IMSC 1.1 we have rh and rw so there is a lossless 1:1 translation from cell.

Andreas: There's more work to be done that's clear.
... To say it is simple, everything IMSC, wherever it is, I see some kind of agreement that this is the long
... term goal, and possibly to come together and discuss it maybe at IBC. We can try to set up a new
... joint meeting at IBC.

Pierre: What's the timeline for submission of EBU-TT Live and EBU-TT Part 1?
... And starting that convergence process.

Andreas: For Part 1 it is different, we haven't started the discussion. For Live it is easier, depends on
... Nigel who has the most work on it.

Pierre: That could be part of what we try to do this year with IMSC.

Frans: To do the transfer, this year?

Pierre: Yes, as well as other additions to IMSC that we have discussed, like embedded fonts.

Nigel: I'm not clear what needs to be done to IMSC though?

Pierre: Exactly, that is why I am optimistic.

Andreas: The live module can be done this year.

Pierre: And make a goal to resolve differences with EBU-TT Part 1.

Andreas: That is more complex, I don't see Part 1 alignment this year.

Pierre: If we are targeting 3 years it would be ideal to do it this year.
... I think it would be different if there had been broad adoption of EBU-TT Part 1.

Frans: I agree. We need a clear understanding of how it is used.

Marco: For example live ingest - how do you translate that to time expressions in a Part 1 document.

Pierre: What we saw yesterday is the bigger issue with live is not the timebase, but the separation of
... flows of audio/video and subtitles.

Frans: But part 1 not live.

Pierre: This will never be used in live?
... What would it be used for?

Frans: Prepared content

Pierre: Why use timecode?

Marco: To synchronise with the playout server.

Frans: Longer discussion!

Andreas: We can schedule for live in W3C, for part 1 we need further discussions and we should have
... this kind of coming together.
... What about this goal to be at IBC.

Nigel: Logistical issue of timing of meetings - TPAC is very close to IBC, I can't be at both probably.

Frans: IBC is where we can collect user input. We need that.

Andreas: We don't necessarily need the people in the room but we need the feedback channeled through us.

Frans: IBC is an opportunity to meet industry people.

Pierre: Do we present results or gather feedback?

Frans: Both, there is not enough time to gather information about timecode usage at IBC

Andreas: A half day workshop would be useful, after the standards people get together.

Pierre: So present results at IBC.

Frans: That would be great.

Pierre: I'm pretty confident that North America is happy to use IMSC for production, I'll hear more in 2 weeks.
... For live I don't know.

Matt: Live is typically 608 originated and delivered straight into EEG or Evertz encoders. It is ripe for evolution.

Pierre: I don't know about live, I'm pretty sure IMSC will be good enough, I've not heard any objection to that.
... Would be good to hear from EBU members.

Frans: It is an extremely early stage.

Andreas: We can cover Live and EBU-TT-D at IBC.
... Then we can start to get out the message about part 1 but I agree with Frans it is more complicated.
... This overall meeting would be good.
... Nigel you say they are close together?

Pierre: The only way I could do TPAC is to be there for a day and a half.

Andreas: For the TTWG we had the idea to move the regular meeting out of TPAC. That would be a question
... if we would move the meeting to Amsterdam.

Pierre: Tokyo would be a lot closer in travel time than Fukuoka!
... By the way I did complain to W3M and they said they don't have IBC and NAB on their calendars!
... They said it would be a good idea.
... The AC meeting is right on top of NAB and TPAC on top of IBC.

Andreas: To sum up we try to move forward offline with EBU-TT Live as quickly as possible and we also
... have some discussion about EBU-TT-D and Part 1 and share info soon.
... Then we propose to both groups to have a domain meeting where we present our results but also a
... face to face meeting that would be joint EBU and TTWG. We have absent TTWG people today so we have
... to propose it.

Pierre: If you're organising meetings at IBC you're not going to make it to TPAC.

Matt: I would go to IBC but not TPAC.

Frans: So hold an industry meeting and a joint meeting both at IBC?

Matt: I'd request immediately before IBC or after, not during.
... So we travel to Amsterdam once and stay.
... If we could do it on the Thursday that would be good for me.
... I will be busy after then.

Andreas: That makes sense.

Matt: I'm usually exhausted by the Tuesday.

Andreas: We need to discuss this in the TTWG.

Pierre: If TTWG is not going to meet at TPAC that would be good to know as early as possible.

Andreas: It would be good to send a message out at IBC.

Matt: Possibly arrange a presentation, say in the innovation presentation area.

Andreas: Those are the things that come to my mind. Anything else we need to discuss or finalise?

Pierre: Is it bad not to attend TPAC?

Thierry: There's nothing mandatory but it is convenient for coordinating with other groups. There are no
... rules about it, it is just for convenience.
... The dates for meeting would be?

Andreas: We need to check it offline.

Pierre: Either Thursday 12th Sep or Tuesday 17th.

Andreas: If nothing else we can close the joint meeting.

Nigel: Thank you everyone, a good morning's work and good discussion, we can adjourn now.
... [adjourns meeting]

Summary of Action Items

Summary of Resolutions

[End of minutes]

Minutes manually created (not a transcript), formatted by David Booth's scribe.perl version 1.154 (CVS log)
$Date: 2019/02/01 12:44:30 $