- From: Nigel Megitt <nigel.megitt@bbc.co.uk>
- Date: Fri, 30 Oct 2015 07:17:01 +0000
- To: Timed Text Working Group <public-tt@w3.org>
- Message-ID: <D259487C.2AD99%nigel.megitt@bbc.co.uk>
Thanks all for another productive day. Minutes for today's meeting can be found at http://www.w3.org/2015/10/29-tt-minutes.html
In text format:
[1]W3C
[1] http://www.w3.org/
- DRAFT -
Timed Text Working Group Teleconference
29 Oct 2015
See also: [2]IRC log
[2] http://www.w3.org/2015/10/29-tt-irc
Attendees
Present
Andreas, Dae, Glenn, Nigel, Phillipe, Pierre, jcdufourd
Regrets
Chair
nigel
Scribe
nigel
Contents
* [3]Topics
1. [4]Charter
2. [5]IMSC & TTML industry feedback, profiles etc.
3. [6]IMSC Issues - proposals review
4. [7]Agenda
5. [8]Joint meeting with Multimodal Interaction Group
6. [9]Charter (revisited)
7. [10]IMSC issues proposals review (continued)
8. [11]IMSC 2 and TTML2
9. [12]TTML2
* [13]Summary of Action Items
__________________________________________________________
<trackbot> Date: 29 October 2015
<scribe> scribe: nigel
Charter
nigel: Our charter runs out at the end of March 2016.
... So we have an opportunity to revise the charter in whatever
way we see fit.
... As a result of the Web & TV IG meeting on Monday I put a
call out for input.
... I received two comments. One was that HTMLCue might be
related to the TimingCue
... being proposed by the Multi-Device Timing CG.
... The other was that we need to include other timing event
models such as SMPTE
... timecode, MIDI etc. which I think we already do in TTML.
... Possibly there's an exception for external event based
triggers.
[14]http://www.w3.org/2014/03/timed-text-charter.html
[14] http://www.w3.org/2014/03/timed-text-charter.html
scribe: Since we have no concrete proposal for HTMLCue to be
defined in this group I think
... we can discount the first comment, and the second comment I
think is also already
... handled by our existing work.
... I haven't had any other proposals for change.
pal: IMSC 2?
nigel: We can already do that with the existing wording, which
is vague enough to
... allow for multiple versions of a spec.
plh: You can keep it as vague as it is now and still do IMSC 2.
pal: We need to update the links from Mercurial to git - that's
editorial!
... The links are in the Deliverables section.
nigel: We should just update them to point to /TR.
pal: And the milestones need to change.
glenn: Can we eliminate the Milestones section?
pal: +1
plh: You'll probably get an objection from the W3C EO.
nigel: Can we rename it "Forecast"?
glenn: That would work.
pal: If we have IMSC 2 do we need a new line in the milestone
section for it?
plh: The goal for milestone is to indicate the likely resource
usage.
atai2: It's important for industry to know this information
too.
plh: You don't need milestones for every deliverable, but
having none would be worrying.
nigel: Can we make the milestones a link to something else that
we maintain?
plh: I believe so, I'd have to double check.
<plh-web> [15]https://w3c.github.io/charter-timed-text/
[15] https://w3c.github.io/charter-timed-text/
plh: That's a link to the online version to your draft charter.
<plh> [16]https://github.com/w3c/charter-timed-text
[16] https://github.com/w3c/charter-timed-text
plh: You have write access to it.
... When we are ready to submit it to the AC I will remove the
write access to it.
nigel: Why not create a branch?
plh: True, I could do that or create a dated version that only
Thierry and I can change.
... At some point I just need a guarantee of stability.
nigel: For the process, when do we need to have this ready?
plh: You need to give me a charter at the end of January.
... Then I take your draft, go to W3C management, and make the
case for spending the
... resources on this WG. That takes ~2 weeks to get approval.
... Then once we secure management approval it is submitted to
the AC for review.
... That takes 4 weeks, per process. Then after that the
Director has 2 weeks to follow up.
... Additional delays could be introduced by review feedback,
if the charter is controversial.
... One strategy for dealing with that is to give the draft
sooner. The other is to indicate
... that it will take a while, and request a ~3 month extension
to the current charter
... while the potential objections are resolved.
... If you don't have a charter ready by the end of January
please let me know because
... you will need an extension to cover the gap.
... One consideration to have, which I can raise as an issue on
github, is which
... license do you want to use for the documents? You can use
the W3C document licence
... that is the one you are currently using, which doesn't
permit document forking and reuse.
... There's another licence available that permits forking as
long as you give attributions.
... It's called the W3C Software and Document licence. It's an
OSI approved licence that
... was modified to include documents.
... You can use whichever one you want. WebVTT will want to use
the Software and Document
... Licence.
glenn: Do we have to choose one for the group?
plh: Only two working groups are using it so far. Other groups
decided not to use the
<scribe> ... new licence. I don't know any reason why you
shouldn't use different licences for
UNKNOWN_SPEAKER: different documents. If you want to do more
complex stuff but not put it in the charter,
... then you could look at the HTML WG charter, which allows
for a discussion during the
... course of the WG to decide later.
... I don't recommend one way in particular.
nigel: Is there a licence specified in the current charter?
plh: No - we only had one licence at the time, so there was no
discussion.
<plh>
[17]http://www.w3.org/2015/10/webplatform-charter.html#licensin
g
[17] http://www.w3.org/2015/10/webplatform-charter.html#licensing
plh: I recommend that you look at the wording of the above web
platform charter as a
... source of inspiration. For example for wording about where
most of the work happens.
... The Decision Policy is an example. Also look at the
Milestones section, which is
... very vague. That WG has 40-50 documents, but only 5 are
listed. That's another example.
... That includes a link to the Specification Status as a
pointer for accurate current data.
nigel: Can we use the same stylesheet for the Charter as the
new one for /TR?
plh: I'll ask but it may not seem important enough to spend a
lot of time on.
... Now that you mention it there was a thread on the AC forum
about a new style for charters.
<plh>
[18]https://lists.w3.org/Archives/Member/w3c-ac-forum/2015OctDe
c/0006.html
[18] https://lists.w3.org/Archives/Member/w3c-ac-forum/2015OctDec/0006.html
pal: So what's our process for developing this?
nigel: Do you want to treat me as the Editor and send PRs?
pal: Okay.
plh: I can do a first pass at changing the links etc.
nigel: The most important thing is the End Date!
plh: Your maximum is 2 years.
nigel: Everyone happy with 2 years?
all: yes
glenn: By the way, 2018 will be the 15th anniversary of this
WG!
plh: The start date will be whenever the Director approves the
charter.
nigel: Right now the only substantive change we have on the
table is the end date.
... Apart from that we're just looking at polishing and
modernising.
plh: In that case you might just want an extension.
nigel: Is there a maximum duration of extension?
plh: I'm not sure - we need to check what the AB is comfortable
with nowadays.
... By the way if you change the document licence for WebVTT
that would be a substantive
... change.
... Although it may not matter because the CG version is
published under the CLA anyway.
... I'll ask Simon what he thinks.
atai2: I wonder about adding deliverables that ease adoption of
our specs like primers etc.
plh: We haven't got enough resources for Recommendations - if
you know someone
... who is willing to do the work on Primers (which I like)
then you can add it.
nigel: We don't need new Notes to be in the Charter.
plh: We got objections in the past when Primers were excluded
from the charter.
... For example the Web Performance Group charter - the absence
of a primer there
... generated a formal objection.
nigel: It's a shame we had to swap the agenda around today to
bring this discussion
... ahead of the industry feedback section. One thing on my
mind is if we should publish
... a document that explains how to stream TTML to avoid us
having the same conversations
... over and over again.
plh: It can be a good technique to get someone outside the WG
to work on primers,
... because they have an external viewpoint. You could ask them
to join the WG for the
... purpose of writing the primer.
... You might want to look carefully at the Scope section, for
example to remove
... anything you know you will no longer need to deliver.
nigel: Looking at Dependencies, do they now need to change?
plh: You want HTML WG and Web Platform WG.
... PFWG is now called something else.
... Leave it to us to update the names.
... The groups in the Liaisons section all still exist.
nigel: We should take Good Standing out?
plh: Correct. Other changes will need to happen to keep up with
the new process.
... Look at the Web Platform charter for Decision Policy. I
don't think it will have a practical change on your operational
mode.
atai2: The Web Annotations group has been considering using
Timed Text. Members
... of that group have mentioned that it may be helpful to meet
up with them. Also,
... the same people are in the Digital Publishing group, so it
may be helpful to meet up
... with them. If there's an opportunity for collaboration then
that would be good.
nigel: I think one WG needs to be responsible for any single
deliverable.
... Do you think that would be a liaison or a dependency?
atai2: Yes, one of those.
nigel: Okay, we need further discussion on that.
IMSC & TTML industry feedback, profiles etc.
atai2: Over the past year we've seen really good uptake of TTML
in the EBU-TT-D profile
... especially because of HbbTV 2.0. For example there's a
prototype implementation
... television from Samsung. I asked if they had any feedback
for the group.
... First they said it is very good, and they would also like
to implement IMSC.
... Also they said they had a good experience working with
people involved in writing
... the specification, and they would like to work with
members.
... They have implemented linePadding and multiRowAlign.
... Another comment is that there is no precise identification
mechanism for IMSC and
... EBU-TT-D even though they are different. We should consider
a precise identification
... mechanism, bearing in mind the option to add a profile
attribute.
... I think the main message is that they need a precise
identification mechanism when
... they work with different profiles. We heard this from
others too, and on the mailing
... list.
nigel: Can you explain the use case more precisely?
atai2: No I can't. For a processor that gets a file they have
to look at the document to see
... if it is EBU-TT-D or IMSC. They have to choose or decide if
they can play it or which
... kind of settings they must apply. I think that's what they
need.
glenn: I can give a couple of use cases too for having some
kind of profile specification
... that is in the document, e.g. the profile attribute. When
landing on a client, to
... determine if the decoder can support it. In the absence of
any profile specification on
... a generic decoder, in the absence of external knowledge,
then it has to use the default
... "dfxp transform" profile. That's safe because no content
will be rejected.
... Another use case is for validation and verification, where
specifying the TTML1
... profile is a hint as to what may have been expected. We
know it is not a content
... profile specification, but it does give some guidance.
jcdufourd: Our use case is that TTML is one of many media that
we want to package
... in MP4 for a segment, or for DASH segmentation, e.g. 2-10s
long segments. We want
... to deal with TTML as with any other kind of media, like a
video gives a precise profile
... indication that's commonly used. It's the same in any
ISOBMFF version not just MP4.
... There's this strange thing that in audio there's the same
problem as in TTML. There's
... no indication without going in deep. dsinger said a couple
of days ago that it's good
... that we only use AAC Low Complexity otherwise noone would
be able to use it.
... That's a similar situation where not having a correct
indication in the stream is a problem.
... Cyril told me that the default profile is unhelpful and
there are multiple ways of
... signalling profile. There's no easy sniffing heuristic. So
all of that makes it difficult
... to use the media TTML, so from our systems point of view if
we have a difficulty using
... one codec then we would use another one.
... Please make it easier for us!
glenn: Of course with WebVTT we don't know which features are
used.
pal: There was an issue filed against IMSC 1 based on feedback
from Cyril. Based on
... that issue I have a proposed resolution. Maybe we could
look at that?
atai2: We need to give clear guidance to implementors, and
maybe to abstract a bit from
... what is correctly written in TTML and find a practical
solution for now.
issue-448?
<trackbot> issue-448 -- Add recommendation for ttp:profile and
ebuttm:conformsToStandard -- pending review
<trackbot>
[19]http://www.w3.org/AudioVideo/TT/tracker/issues/448
[19] http://www.w3.org/AudioVideo/TT/tracker/issues/448
pal: Presents proposed edit to the #profile feature.
... One problem is that ttp:profile attribute only supports a
single value in TTML1, so you
... cannot signal dual conformance with EBU-TT-D and IMSC 1
Text profile.
... Also EBU-TT-D prohibits use of ttp:profile.
... So the proposal is to use ttp:profile="[imsc profile
designator]" for IMSC 1, and
... omit ttp:profile but include ebuttm:conformsToStandard for
EBU-TT-D.
jcdufourd: And ttp:profile element is prohibited?
pal: Yes. The reason for that is that the element is used to
specify a profile specification
... in-band, and IMSC 1 does not specify such a profile.
jcdufourd: Why 'SHOULD' not 'SHALL'?
pal: There will still be instances where it is not present for
a while. It would be extreme
... to reject otherwise conformant documents just because they
do not include the ttp:profile attribute.
... I'm happy to put SHALL though.
... The spec also says that if a document is not conformant
then the behaviour is not defined.
jcdufourd: So if neither of these things is present then what
would you do?
pal: If I were building an implementation then I would force
the "user" (broadly) to tell
... the implementation what the file is.
jcdufourd: That matches what nigel recommended privately.
nigel: Yes, I said the same thing.
glenn: In TTML1 we say that the processing environment can
infer a profile in the absence
... of any information in the document, and if that fails then
there's a default minimum profile.
jcdufourd: The use case in which inferring is not good is in
the live situation where you
... stream a presentation with live subtitles and you have no
idea what will happen in
... the future so you cannot sniff ahead.
glenn: Since TTML does not define a streaming format what
people have done is to
... chunk into separate small TTML files, each of which can
specify the profile. If they
... change this between documents, then that would be
interesting.
... I would prefer 'SHALL' also. I wonder what would happen
with CFF content. In that
... case I guess it would be implementation dependent since
those existing files would
... not be valid.
pal: This can't be pain-free, but we should minimise the pain!
nigel: What about EBU-TT-D documents that don't support imsc or
do - what should we say about ebuttm:documentConformsToStandard
carrying one or two values?
group: [discusses this] It's actually covered.
atai2: The recommendation is to have both values signalled in
...conformsToStandard.
... Maybe we should clarify that if imsc is omitted from
...conformsToStandard then it should be considered
non-compliant with IMSC.
pal: If we write SHALL then it implies that implementations
give up if they don't see
... exactly what's here, but the proper behaviour is that
there's an opportunity to specify
... the profile externally.
jcdufourd: What I understand is either 1) IMSC or 2) EBU-TT-D
or 3) External profile or
... 4) I can trash it. The flavour is that there is a SHALL.
pal: So an out of band specification doesn't break the SHALL?
jcdufourd: Correct. If that's the understanding then we're
good.
glenn: Is there any reason you would have to trash it? You
could just use the default default.
jcdufourd: The thing is that our way of doing it in systems is
to use the maximum profile.
nigel: Actually that's the same thing.
glenn: If you choose the maximum then you're being the most
conservative about what
... you require from the profile. If you were talking about
content profiles then you'd be
... saying that the document uses all the features but the
processor does not need to.
jcdufourd: I don't see a difference from the systems
perspective.
glenn: The difference is that you wouldn't necessarily throw
the document out.
pal: You can set an out of band policy for a specific profile.
glenn: It's fair to be able to do that.
pal: If I find a document with conformsToStandard="ebu-tt-d"
but not Imsc1 then that's
... fine and the logical conclusion is that I can't assume that
it's an IMSC 1 compliant document.
nigel: That would handle the corner case of say an EBU-TT-D
document that's UTF-16, say.
ack
<Zakim> glenn, you wanted to ask if ebu will remove the
restriction on ttp:profile
glenn: That goes back to my question. If you have a generic
IMSC or TTML processor and
... no profile attribute is present then it can handle it, and
if one is present than it can
... handle it but if it is prohibited then that introduces the
problem of having to guess.
... You really want both ttp:profile and conformsToStandard to
be present, because
... an IMSC processor that does not know about
ebuttm:documentConformsToStandard
... would not know what to do.
... It's hostile to interoperability to prohibit ttp:profile.
What's the chance of EBU
... changing it?
atai2: Realistically, the spec is closed apart from errata.
glenn: In a future version?
atai2: I agree it's better to have it all in one place, but as
pierre said you cannot signal
... multiple profiles to be present. It would have to be
discussed in EBU. At the moment
... we need to find a solution for current documents. This from
my side is a good solution.
nigel: I agree this is a proposal that will work pragmatically.
The thing we haven't
... considered is adding a new thing to IMSC that could find
its way into TTML later
... that permits multiple profiles to be indicated, perhaps
using the profile registry
... short codes.
pal: I'm happy that this current proposal satisfies the issue
that's been raised.
nigel: Agreed. Adding a new thing would give e.g. EBU something
to discuss that may deal with their issues.
atai2: We can also look at how the profile registry document
can support this.
pal: I'd like to close this issue because it addresses the use
case and then look at that separately.
atai2: Additionally, did we agree that it is possible to signal
the profile externally without
... using this?
pal: Yes. The profile designator can be specified out of band.
nigel: So are we adopting this proposed #profile text unchanged
or are we changing
... any of the SHOULDs to SHALLs?
jcdufourd: If the out of band signalling is meaningful and not
rare then you can leave the SHOULDs in.
glenn: I'm fine with leaving it as SHOULD. I think there needs
to be language somewhere
... that says if neither are present then what behaviour is
expected or any inference should
... be made.
pal: That's the implementor's problem.
atai2: Undefined.
nigel: We can be a bit friendlier and add a note to say that
out of band signalling is expected
... if the SHOULDs are not honoured.
glenn: So there's no inference?
nigel: The inference is if there's no in-band and no
out-of-band signalling then it's not
... related to this specification!
pal: I would also say authoring tools should never be able to
create documents that don't
... signal the profile.
jcdufourd: +1
pal: I don't like to add a Note here about out of band
signalling. I don't think it's helpful to implementors.
nigel: Okay but we haven't dealt with the feedback 'what to do
if there's no in band signalling'?
pal: I think that's something to add to an external usage note
like the profile registry
... document.
jcdufourd: My name is Jean-Claude Dufourd. I work at ParisTech,
alongside my colleague Cyril who cannot be present today.
Shingo: My name is Shingo Mine from Mitsubishi.
atai2: I propose that in the profile registry document we add a
recommendation on where
... to find profile information in-band as well as out-of-band,
to make it clear for all
... implementors where they can find this information.
<scribe> ACTION: atai2 Propose to mdolan this addition to the
profile registry document. [recorded in
[20]http://www.w3.org/2015/10/29-tt-minutes.html#action01]
<trackbot> Created ACTION-445 - Propose to mdolan this addition
to the profile registry document. [on Andreas Tai - due
2015-11-06].
nigel: Let's break - back at 1100.
close issue-448
<trackbot> Closed issue-448.
IMSC Issues - proposals review
issue-410?
<trackbot> issue-410 -- Constraints on #linePadding and
#multiRowAlign -- pending review
<trackbot>
[21]http://www.w3.org/AudioVideo/TT/tracker/issues/410
[21] http://www.w3.org/AudioVideo/TT/tracker/issues/410
group: reviews text on #linePadding and #multiRowAlign in the
current ED.
glenn: You can not use the word apply in this way.
ebutts:linePadding has no effect on region.
... The only place it has an effect is on a span.
atai2: The definition of this does describe its application
relative to a containing block.
... The meaning is exactly as you describe. The apply section
is about the semantic
... application of this concept to elements in TTML.
glenn: Another example is foreground colour, which can be
specified on any content
... element type but it only applies to text elements inside a
span.
... Similarly extent and origin apply to region, which was
misconstrued because we
... didn't say that there's no semantic when it is on an
element from which there is no
... inheritance down to region.
atai2: That's 100% the meaning we intend.
glenn: It only applies to span though.
atai2: But the EBU-TT-D definition is for application to parent
elements.
... There seem to be 2 issues. One is not related to IMSC,
being the definition of the feature.
... The other is the conformance language, which is the issue
we have to deal with here.
nigel: We have to consider "applies" relative the feature as
defined, which is on the block element.
glenn: I have to consider the equivalent functionality in
TTML2.
nigel: I disagree - you can treat that independently.
...
glenn: Line areas are all descendants of the p element so it
makes no sense to apply
... linePadding to region, body or div.
... [draws diagram showing block, line and inline areas] Line
areas are only generated
... by block areas associated with p elements.
nigel: Can we resolve by saying that #linePadding applies to p
and is inherited?
group: okay
pal: let me add that...
issue-410: [meeting 2015-10-30] agreed to say for #linePadding
and #multiRowAlign that they shall be applied to p and that
they shall be treated as inheritable.
<trackbot> Notes added to issue-410 Constraints on #linePadding
and #multiRowAlign.
nigel: Also worth noting that the other raised problem was
fixed in issue-450.
issue-410: [meeting 2015-10-30] Also note that issue-450 was
resolved so the conformance language problem no longer applies.
<trackbot> Notes added to issue-410 Constraints on #linePadding
and #multiRowAlign.
issue-411?
<trackbot> issue-411 -- "shall be inherited" on #multiRowAlign
-- pending review
<trackbot>
[22]http://www.w3.org/AudioVideo/TT/tracker/issues/411
[22] http://www.w3.org/AudioVideo/TT/tracker/issues/411
issue-411: [meeting 2015-10-30] See also notes on issue-410.
<trackbot> Notes added to issue-411 "shall be inherited" on
#multiRowAlign.
close issue-410
<trackbot> Closed issue-410.
close issue-411
<trackbot> Closed issue-411.
issue-423?
<trackbot> issue-423 -- Use of proprietary, non-open source
fonts as reference fonts -- pending review
<trackbot>
[23]http://www.w3.org/AudioVideo/TT/tracker/issues/423
[23] http://www.w3.org/AudioVideo/TT/tracker/issues/423
glenn: The switch to open source fonts is acceptable to me.
pal: I also verified that the width is equivalent.
glenn: Is there any need to have the OR?
pal: The reason I kept that is that if you do have the non-open
fonts available you should
... not be penalised, so both should be allowed.
glenn: I would remove the Note in Appendix A too.
pal: I would agree with that - I don't think it is particularly
useful.
... I'll remove it right now.
close issue-423
<trackbot> Closed issue-423.
issue-450?
<trackbot> issue-450 -- IMSC1 Does Not Require Any
Feature/Extension be supported by a Processor -- pending review
<trackbot>
[24]http://www.w3.org/AudioVideo/TT/tracker/issues/450
[24] http://www.w3.org/AudioVideo/TT/tracker/issues/450
nigel: I think we covered this yesterday?
pal: I've now done the commit based on the preview.
... In §4 Conformance I added an i.e. phrase to address Glenn's
comment.
glenn: I don't like seeing "i.e." in conformance language but I
won't object to it.
nigel: Is there an easy quick shortcut to fix that?
glenn: I'll compose something...
<glenn> SHALL implement presentation semantic support for all
features designated as "permitted" by the profile, modulo any
additional constraints on features as specified by the profile;
pal: Can we use "subject to" instead of "modulo"?
<glenn> SHALL implement presentation semantic support for all
features designated as "permitted" by the profile, subject to
any additional constraints on features as specified by the
profile;
group: looks good
glenn: The same change applies to presentation processor and
transformation processor.
<glenn> SHALL implement transformation semantic support for all
features designated as "permitted" by the profile, subject to
any additional constraints on features as specified by the
profile;
glenn: Are there any features that are not "permitted" where
some support is required?
pal: No we only have permitted and prohibited.
nigel: Now that we've done that wouldn't it make sense to add
"permitted" to #lineBreak-uax14?
... and add a note that it has no syntactical impact.
group: [discusses the pros and cons of this]
nigel: Consensus achieved on leaving #lineBreak-uax14 as is,
noting that processor implementors may have to be extra careful
to note
... that implementation of that feature is required even in the
absence of the "permitted" keyword.
close issue-450
<trackbot> Closed issue-450.
issue-450: [Meeting 2015-10-30] Change implemented above as
agreed.
<trackbot> Notes added to issue-450 IMSC1 Does Not Require Any
Feature/Extension be supported by a Processor.
Agenda
nigel: We're visiting Multimodal Interaction in 2F. 204 at
1300, returning here at 1315.
... [adjourns for lunch]
close issue-452
<trackbot> Closed issue-452.
issue-452: [Meeting 2015-10-30] Change agreed as above.
<trackbot> Notes added to issue-452 The ttp:profile element is
permitted to conflict with the IMSC 1 profile.
Joint meeting with Multimodal Interaction Group
nigel: [discussion of TTML example in EMMA 2.0 spec, and
further potential use cases]
Charter (revisited)
glenn: There may be a use case we haven't covered for HTMLCue.
nigel: Maybe if people want us to develop something in this
group then we need to add
... something to the Charter.
... It can be generic, like "A mechanism for displaying
arbitrary HTML with cue timings"
group: [general support]
IMSC issues proposals review (continued)
issue-418?
<trackbot> issue-418 -- semantics of ttp:aspectRatio is
ambiguous -- pending review
<trackbot>
[25]http://www.w3.org/AudioVideo/TT/tracker/issues/418
[25] http://www.w3.org/AudioVideo/TT/tracker/issues/418
glenn: My query is if aspect ratio and extent are both
specified then which one applies?
nigel: Aren't you assuming square pixels?
glenn: No I'm not!
pal: Let's look at the examples.
glenn: My first question is what is the relationship between
the aspect ratio and the
... storage aspect ratio implied by the extent?
pal: None. extent describes the root container logical
coordinates, e.g. 400px 300px.
glenn: That can be used to derive the storage aspect ratio.
... I agree that it only matches the Display Aspect Ratio if
the pixels are square.
... So aspectRatio tells you the display aspect ratio. And you
can derive the pixel aspect
... ratio from that?
pal: That's right.
glenn: Can you add a note to say that aspect ratio means
display aspect ratio?
pal: Sure. Of the root container.
glenn: That's right.
... And that you can derive a storage aspect ratio...
pal: Can you draft that in IRC?
<glenn> Note: The ittp:aspectRatio parameter effectively
defines the intended display aspect ratio (DAR) of the root
container, while the tts:extent style property on the root
element effectively defines the intended storage aspect ratio
(SAR) of the root container.
issue-418: [Meeting 2015-10-30] Proposed note: Note: The
ittp:aspectRatio parameter effectively defines the intended
display aspect ratio (DAR) of the root container, while the
tts:extent style property on the root element effectively
defines the intended storage aspect ratio (SAR) of the root
container.
<trackbot> Notes added to issue-418 semantics of
ttp:aspectRatio is ambiguous.
nigel: Does that resolve it Glenn?
glenn: Yes, that would resolve this issue.
... I was previously interpreting it as storage aspect ratio,
which would have been a conflict.
pal: I've updated the document, just adding the changelog to
the issue.
close issue-418
<trackbot> Closed issue-418.
issue-449?
<trackbot> issue-449 -- Needs XSD Schema for IMSC Extensions --
pending review
<trackbot>
[26]http://www.w3.org/AudioVideo/TT/tracker/issues/449
[26] http://www.w3.org/AudioVideo/TT/tracker/issues/449
pal: I added XSDs
... I was inspired by Glenn's contribution but changed it. The
end result is identical.
... I changed it to define types inline rather than in a new
namespace.
glenn: It turns out to be important for enumeration types when
you generate a JAXB
... binding. If you don't have a top level definition then it
won't generate a type. The
... code in TTV that supports IMSC has to have them as separate
definitions.
pal: But there's no enumeration here though. Just native
booleans.
glenn: I just double checked and it's fine. The only JAXB type
generated was the alt-text element that is defined as a new
element in this.
nigel: Having a look at the schemas... Can't view the files.
pal: I need to update the links to change raw-file to file so
you can navigate the directory.
glenn: And these are not normative so we can change the
schemas?
pal: That's right. Specifically non-normative.
glenn: Notice that they do not include either the SMPTE or
EBU-TT-D entities. In TTV
... I had to add this as well as the SMPTE and EBU-TT-D
features, which I had to subset.
pal: The links are relative, so I don't know how to fix them.
This will probably go away when we switch to github.
issue-449: [Meeting 2015-10-30] Group happy with schemas - can
close.
<trackbot> Notes added to issue-449 Needs XSD Schema for IMSC
Extensions.
close issue-449
<trackbot> Closed issue-449.
issue-451?
<trackbot> issue-451 -- #visibility-inline should be limited to
text profile -- pending review
<trackbot>
[27]http://www.w3.org/AudioVideo/TT/tracker/issues/451
[27] http://www.w3.org/AudioVideo/TT/tracker/issues/451
<scribe> ACTION: pal Go through every feature and make sure
that they take into account any dependent features. [recorded
in [28]http://www.w3.org/2015/10/29-tt-minutes.html#action02]
<trackbot> Created ACTION-446 - Go through every feature and
make sure that they take into account any dependent features.
[on Pierre-Anthony Lemieux - due 2015-11-06].
close issue-451
<trackbot> Closed issue-451.
issue-453?
<trackbot> issue-453 -- Clarify relation between forcedDisplay
and visibility in intro statement -- pending review
<trackbot>
[29]http://www.w3.org/AudioVideo/TT/tracker/issues/453
[29] http://www.w3.org/AudioVideo/TT/tracker/issues/453
close issue-453
<trackbot> Closed issue-453.
pal: There are some issues still open on HRM, which are all
related.
... The difference between GCpy and Ren is the performance
improvement from glyph reuse.
... Based on the discussion yesterday glyphs cannot always be
reused for a particular code point,
... because a code point has multiple variant glyphs. The idea
is to reduce the performance
... improvement that is predicted based on script, because some
scripts have more
... variant glyphs than others.
... The distinction for CJK is only outside the glyph buffer
for text rendering.
glenn: So you might have different values of GCpy based on
script?
pal: Right. However as you pointed out that there are different
ratios for different scripts.
glenn: The question is if it is worth changing the model to
take into account Indic.
... In Indic there are other steps too, like reordering. I
would say on average treat Indic
... the same as Arabic for the multiplier. I suggest using two
terms, "simple" and "complex"
... which are commonly used in the industry. Unfortunately
there is no Unicode term for
... this. There is an enumeration of scripts, and there are ISO
script codes.
[30]https://en.wikipedia.org/wiki/ISO_15924
[30] https://en.wikipedia.org/wiki/ISO_15924
glenn: You could enumerate the script codes and enumerate them
as simple or complex.
... I'd suggest varying the HRM based on those two labels, and
I can tell you which scripts
... fall into which category offline.
<scribe> ACTION: glenn Send Pierre the list of which scripts
are simple and which are complex. [recorded in
[31]http://www.w3.org/2015/10/29-tt-minutes.html#action03]
<trackbot> 'glenn' is an ambiguous username. Please try a
different identifier, such as family name or username (e.g.,
gadams, ggoldste).
<scribe> ACTION: gadams Send Pierre the list of which scripts
are simple and which are complex. [recorded in
[32]http://www.w3.org/2015/10/29-tt-minutes.html#action04]
<trackbot> Created ACTION-447 - Send pierre the list of which
scripts are simple and which are complex. [on Glenn Adams - due
2015-11-06].
action-432?
<trackbot> action-432 -- Nigel Megitt to Nigel, pal, andreas
and frans to hook up re changing imsc 1 ebu-tt feature
references to point to tech3350 v1.1 -- due 2015-10-15 -- OPEN
<trackbot>
[33]http://www.w3.org/AudioVideo/TT/tracker/actions/432
[33] http://www.w3.org/AudioVideo/TT/tracker/actions/432
pal: We've made a decision how to progress.
close action-432
<trackbot> Closed action-432.
action-437?
<trackbot> action-437 -- Nigel Megitt to [imsc1] propose new
wording for intro sentence to itts:forceddisplay -- due
2015-10-22 -- PENDINGREVIEW
<trackbot>
[34]http://www.w3.org/AudioVideo/TT/tracker/actions/437
[34] http://www.w3.org/AudioVideo/TT/tracker/actions/437
close action-437
<trackbot> Closed action-437.
pal: Can we talk about set in IMSC?
... I got an email asking how to express in IMSC 1 a real world
example where a title
<pal> pal: use case is " A real world example is when a title
or graphic suddenly appears in the video, a good real-time
captioner will send a command to move the existing accumulated
captions to another part of the screen so that they don’t
obscure the graphic. How do we represent that in IMSC? We
thought of some options"
pal: The two options they could think of were a) duplicate the
text in a different region.
... b) use set to move the origin of the region.
... They asked for any other ideas.
glenn: Are they asking for smooth scrolling?
pal: No just a jump.
glenn: Both of those would work.
nigel: I want to know why they need to know if the text is the
same text as opposed to
... just being interested in the appearance being correct.
glenn: They could define some of their own metadata to identify
the content, if they use the first solution and want to keep
the semantic meaning.
nigel: I'd observe that in EBU-TT-D you can't use set, but if
they don't need that then they can use set.
... The usage of a content identifier is useful not just for
spatial movement/deduplication
... but also temporal, in case the same text appears in
consecutive TTML documents in
... a streaming scenario. If you want to push the text into a
screen reader or TTS engine
... then a content identifier allows the implementation to
avoid speaking the same words
... multiple times.
pal: Has anyone solved this?
nigel: I'm told that the streaming solutions for VTT don't have
this effect.
atai2: This has been discussed in other groups also.
glenn: It turns out that Lambda cap does have an id called
screen number but it is not
... consistently used, so it's hard to maintain that metadata.
pal: So the downside of using set is that in general it would
not deal with semantic linking
... temporally.
nigel: Yes.
atai2: Is there a requirement for an id independent from xml:id
that makes it possible
... to have a linkage between content elements.
nigel: xml:id is tightly specified to have single document
uniqueness. If you want something
... that crosses documents you need to define something else.
... Is there a generalised ID scheme for content across and
within XML documents?
glenn: Not that I'm aware of.
... I would imagine in the Text Encoding Initiative there is
markup that's defined
... for this purpose.
pal: Okay thanks I'll respond based on that.
IMSC 2 and TTML2
pal: I have a draft set of proposals for what to include from
TTML2.
... One worry is the condition construct.
glenn: You could subset it to parameter based conditions only,
to satisfy forcedDisplay.
pal: I was hoping to have a wider session to think about the
whole condition construct.
glenn: The existing condition language is less complex than
other similar languages like
... in CSS. I had in mind boolean expressions, but there
doesn't seem to be any point
... in not having relative arithmetic too with less than or
greater than for example.
... We also have media queries which I had incorporated into
that condition language.
... They are commonly used in CSS and in mobile contexts.
... I thought it would be useful to design TTML content that
could work on devices where
... it chooses different video formats and resolutions based on
the screen aspect ratio
... and resolution. You could create different files for each
context; another is to use
... condition.
nigel: This is an interesting moment to show a demo of some
work done by some of
... my colleagues in BBC R&D. [shows demo of text size vs time
variance]
glenn: You could do this by duplicating content and using the
condition system to
... select content based on font size range.
nigel: My thinking was to use an earliest and latest time
expression semantic, in other
... words extending the time expressions.
pal: Why is this not purely a client feature?
nigel: You need to have some authorial input because you may
want to restrict the
... earliest or latest appearance time of e.g. words.
pal: You could add metadata to add earliest/latest semantics.
glenn: You could also conditionalise line breaks.
atai2: Can we collect some more detailed use cases for this?
glenn: If you had a tag that would clear a region and you could
conditionalise that then
... it could help with this.
pal: In general it would be better to list the use cases to
meet here for this conditional
... language and to accommodate what you have shown, to see
what more would be
... needed to meet these requirements, and to decide if we
should cover those use cases.
dae: That's called Responsive Subtitles?
nigel: Right.
dae: Another use case is to change the colour scheme.
atai2: Also a request to support better customisation in TTML
directly would be a part of it.
... Users often want support for customisation.
nigel: One problem we have is how to make it possible for
implementations to be
... processor spec compliant and also accessibility guidelines
compliant.
glenn: We had a thing a while ago called Dynamic Flow
Processing Model.
... I spent a long time trying to come up with a rational model
for this!
<glenn>
[35]http://www.w3.org/TR/2010/CR-ttaf1-dfxp-20100223/#dynamicFl
owModel
[35] http://www.w3.org/TR/2010/CR-ttaf1-dfxp-20100223/#dynamicFlowModel
glenn: [explains]
nigel: I don't propose to bring that back!
pal: I want to clarify why I don't think it's necessary for
implementations to offer the
... user to change colours etc. The parallel is televisions,
which receive a very specific
... coding format, with a colour space, a reference decoder
etc. but of course the TV
... applies other processing like denoising, vivid mode etc.
There's a clear parallel, where
... you can have a core conformant processor and yet the
application can choose to apply
... additional processing like choosing fonts.
atai2: I think possibly the author may need more influence than
that to follow
... accessibility, with data in the document not just in the
implementation. The client
... side does not know about the content, only the author does.
pal: There's a regulatory regime that players have to meet
already, in the US.
nigel: It would be nice if implementations could meet
regulations and be spec compliant.
pal: They already are. The spec does not prevent customisation.
nigel: If a user chooses a font size then except in the one
case that they choose what is
... in the file then the processor can either meet regulations
or the spec.
atai2: Nevertheless it would be useful to provide guidance on
how customisation could
... be provided.
pal: I am not aware of inconsistencies between specification
and regulations.
glenn: I can't see any language in TTML1 that allows for user
overrides, but that would
... have been the intention all along. We just assumed that it
would happen. I don't
... think that comment has ever been raised before.
... We could make an errata to state that the processing
environment can override
... presentation requirements as requested. We should also
think about that in TTML2.
... There are a couple of pending review items on TTML...
TTML2
issue-224?
<trackbot> issue-224 -- Support text placement in 3D coordinate
spaces (not zIndex compositing). -- pending review
<trackbot>
[36]http://www.w3.org/AudioVideo/TT/tracker/issues/224
[36] http://www.w3.org/AudioVideo/TT/tracker/issues/224
glenn: I applied Pierre's patch. I have since learned that some
edits may be needed.
pal: I corrected this in the branch.
glenn: Unless it's ready I won't reapply it.
... Looking at the sentence that begins "tts:disparity" please
could you edit it so it does not begin with a keyword?
... I didn't spot the latest email and just applied the
previous patch.
... Maybe we can do that offline.
... Also I edited it a little bit to remove the reference to
stereoscopy.
... The changes made between 25th and 27th were just language
changes?
pal: yes.
glenn: Okay I'll fix that up.
issue-322?
<trackbot> issue-322 -- Formula for dropNTSC time expressions
is incorrect. -- pending review
<trackbot>
[37]http://www.w3.org/AudioVideo/TT/tracker/issues/322
[37] http://www.w3.org/AudioVideo/TT/tracker/issues/322
glenn: We had from June 2015 an email from Charles Ritchea
pointing out an error
... in Appendix N.3. He points out that the floor() is in the
wrong place.
... And he shows the fix for the dropPAL case also.
... I recently verified these numbers with some actual test
content and looking at the
... exact generated timing, and these corrections produce the
right results. I've fixed
... these in two places. There's an errata update for TTML1:
<glenn>
[38]https://dvcs.w3.org/hg/ttml/raw-file/2cd6f1c06e57/ttml1/spe
c/ttml1-errata.html
[38] https://dvcs.w3.org/hg/ttml/raw-file/2cd6f1c06e57/ttml1/spec/ttml1-errata.html
nigel: Looks like you haven't pushed the same changes to TTML2
yet.
glenn: So I'd like approval to publish the new errata and once
I've verified it in TTML2 to close that out.
pal: Where did you find the formulas?
glenn: Originally I derived them from some specifications in
SMPTE specs that defined
... the drop frames very precisely, but it looks like I got it
wrong.
... I can't remember the exact specs.
issue-322: [Meeting 2015-10-30] Group happy with proposed
errata text at
[39]https://dvcs.w3.org/hg/ttml/raw-file/2cd6f1c06e57/ttml1/spe
c/ttml1-errata.html
[39] https://dvcs.w3.org/hg/ttml/raw-file/2cd6f1c06e57/ttml1/spec/ttml1-errata.html
<trackbot> Notes added to issue-322 Formula for dropNTSC time
expressions is incorrect..
nigel: I want to use the last few minutes of Glenn's time to go
through some timing
... pictures following on from what we discussed in Las Vegas.
... [Document time vs media time picture]
... [Constrained by some external process] - the constraint
does not offset the contents.
glenn: I need to check that Root Temporal Extent does not
always start at zero.
... The constraint of active window fits the SMIL concept.
group: moves on to the subject of media offsets
glenn: The proposal is to move a media offset to metadata.
nigel: Yes - I think the most useful thing is to indicate the
beginning of the programme
... in the document's timeline.
glenn: Wouldn't that have the same values as mediaOffset?
nigel: Well maybe but I think the meaning is clearer.
glenn: I'd also like to conditionalise based on ttm:item
nigel: Why not do it based on any extension thing like
myextension:myparameter attribute?
glenn: Well you could do that too.
group: discussion of media time and whether it begins at zero
or not.
pal: We discussed this in the context of IMF and the
overwhelming consensus was not
... to permit negative media times.
<scribe> ACTION: nigel put timing diagrams somewhere they can
be reviews [recorded in
[40]http://www.w3.org/2015/10/29-tt-minutes.html#action05]
<trackbot> Created ACTION-448 - Put timing diagrams somewhere
they can be reviews [on Nigel Megitt - due 2015-11-06].
nigel: Well it's time for a break, and we've done pretty much
everything we wanted to,
... and some folk have to catch their flights, so let's end the
meeting here.
... [adjourns meeting] Thanks everyone!
Summary of Action Items
[NEW] ACTION: atai2 Propose to mdolan this addition to the
profile registry document. [recorded in
[41]http://www.w3.org/2015/10/29-tt-minutes.html#action01]
[NEW] ACTION: gadams Send Pierre the list of which scripts are
simple and which are complex. [recorded in
[42]http://www.w3.org/2015/10/29-tt-minutes.html#action04]
[NEW] ACTION: glenn Send Pierre the list of which scripts are
simple and which are complex. [recorded in
[43]http://www.w3.org/2015/10/29-tt-minutes.html#action03]
[NEW] ACTION: nigel put timing diagrams somewhere they can be
reviews [recorded in
[44]http://www.w3.org/2015/10/29-tt-minutes.html#action05]
[NEW] ACTION: pal Go through every feature and make sure that
they take into account any dependent features. [recorded in
[45]http://www.w3.org/2015/10/29-tt-minutes.html#action02]
[End of minutes]
__________________________________________________________
Minutes formatted by David Booth's [46]scribe.perl version
1.140 ([47]CVS log)
$Date: 2015/10/30 07:01:59 $
__________________________________________________________
[46] http://dev.w3.org/cvsweb/~checkout~/2002/scribe/scribedoc.htm
[47] http://dev.w3.org/cvsweb/2002/scribe/
Scribe.perl diagnostic output
Received on Friday, 30 October 2015 07:17:39 UTC