Meeting minutes
This meeting
Gary: Today we have IMSC 1.2 - the APA WG HR comments, and then the ARIB incoming
… liaison issues that have been opened.
… Followed by TTML2 2nd Ed IR which includes one discussion of tests.
Nigel: That one we can strike off, it's done.
Gary: Then just the IR piece. And are there any other out of band items?
group: [none]
Gary: Then no further business.
IMSC 1.2 APA WG HR comments
Nigel: I've been in touch with APA WG and Philippe. You may have seen there's been
… some movement on the issues this week.
… Hopefully they will all be done this week.
Pierre: On issue #519, we ought to follow up directly with Janina and Gottfried after this
… meeting. My concern is we will be back to another long cycle so you (Nigel) or I should
… follow up with Janina about this particular issue.
Nigel: Okay I'm happy to do that.
Pierre: Hopefully Gottfried will be satisfied with the pull request, I expect it because it is
… factual. We should talk about #520 a little bit.
APA WG comment: Requested Additional WCAG 2.1 References imsc#520
github: https://github.com/w3c/imsc/issues/520
Pierre: The APA suggested that we add a specific reference to the guideline on non-text
… contrast. We already mention the matching reference on text contrast.
… What I don't understand is, by definition, all content in IMSC is text so I don't see
… why this would ever apply. They mention emojis, but where do you draw the line
… between emoji and kanjis. They say anyway you should reference the section.
… We could do it anyway and not care.
… I don't see why it would ever apply in IMSC, generally.
Nigel: My view is we should just add the reference as requested, for two reasons.
… First reason is it doesn't harm.
… Second reason is that although our intent is for image profile to have images that
… mainly contain text, it does not have to be used for that, and any image content
… can be used. Even if the images do contain text, they may contain other visual content
… for which the guideline does apply.
Pierre: On the first point I think it is actually harmful because for
… some reason the non-text contrast is different than the text contrast.
… You wouldn't want kanji and non-kanji to have different contrast in the same sentence.
… On the second point, of course people can do whatever they want, but IMSC is really
… intended for timed text. We say that if there is an image profile document there ought
… to be a text profile equivalent.
… We don't have to account for all things people may want to do.
… I don't see why an image should have a different contrast on screen than text.
… Unless we can point to a specific example within the scope of IMSC for which
… this success criterion may apply I don't think we should include it.
Nigel: Which guideline requires the higher contrast ratio, text or non-text?
Pierre: Text and images of text, 4.5:1, as opposed to 3:1 for non-text contrast.
Nigel: Then it is a no-op - if you meet the 4.5:1 ratio then the 3:1 guideline is exceeded.
Pierre: Yes, we could note that.
Nigel: So we make it a blanket requirement to meet the contrast requirements of both
… text and non-text WCAG criteria.
Pierre: Everything that's relevant to IMSC is in 1.4.11, it is for UI and graphical objects.
… It is not germaine. The one that is really relevant is 1.4.3. All that people need to care
… about is that one.
Andreas: Apologies for my disinformation if I have not got the complete case fully understood.
… Did I understand correctly that they asked for, e.g. for an emoji, that this may be
… considered as image and therefore could have a different contrast than the surrounding
… text only content?
Pierre: I don't really understand @gzimmermann's answer.
Andreas: We work a lot with emojis in text and we definitely wouldn't have a different
… contrast for the emoji against the background than the surrounding text.
… It would be super difficult to arrange and test.
… It would be distracting if there are multiple emojis with different backgrounds
… applied.
Nigel: Are they talking about foreground-background contrast only or contrast within
… the e.g. emoji itself.
… Answering my own question, in the link to understanding non text content, they show
… graphical images and explain that the contrast within the symbols is included.
Pierre: I read 1.4.11 as being targeted at UIs not text that humans read.
Andreas: Yes I fully agree with Pierre because all text needs to comply so
… you would need to combine this with graphic like objects everywhere they may be
… combined with text.
Nigel: I'm a bit worried that by not taking the suggestion we may be preventing something
… that we would like to permit, i.e. that it relaxes the contrast requirement within an emoji
… symbol, so allowing more flexibility.
Pierre: The text contrast requirement is for the border of the glyph compared to the background.
Nigel: Okay but what about within that border?
Andreas: I have an example where there is a black opaque background and yellow text,
… and then you add an emoji that is yellow on the outside. The outline of course is black,
… so the contrast between the border and the background is very low. This is a very
… typical use case. We do this!
Nigel: Good point!
… I don't have an answer for that.
… Just heading back to the comment, as Pierre has noted, the images are images of text,
… so where that applies the 4.5:1 rule applies.
… I think for your example Andreas perhaps we need a note to say that there must be
… enough contrast between the image on the inside of the border and the background.
Andreas: You need a good automated way to calculate what background colour this applies to.
… You would not really recommend a different background colour for an emoji compared
… to the rest of the text.
Cyril: I wonder how this would apply to the use case you presented some time ago Nigel
… where you had the Twitter logo inline with the text?
Nigel: So if the Twitter logo didn't have 4.5:1 to the background, what would you do?
Cyril: Yes, I don't know, is it text?
Nigel: I guess you'd use an alternate version of the image to achieve the contrast ratio.
Cyril: In simple fonts you have one colour per glyph, but for an image there are more colours
… so you have to have different rules. If you store the image in a font stored just like text
… the same text requirement would work.
… Is that true?
Nigel: I have a feeling that you can put colours in fonts now.
Cyril: Yes I meant simple fonts without colours.
… Can we say that the 4.5:1 requirement applies to content inside a font even if it is not
… text only when that text is single-coloured?
Pierre: I don't claim to have the definitive answer but if you look at success criterion 1.4.3,
… there's a hint about why it exists, and an exemption for large text, where the contrast
… ratio is at least 3:1.
… One component of the decision about what contrast ratio seems to relate to size.
… I don't know what "large" is defined as being relative to.
… Trying to bring this to a point, my basic contention is that 1.4.3 is what literally applies
… to IMSC, in general.
… I think we can simply say that IMSC is not intended for generic UI elements and overlays,
… therefore this is the section that applies.
<cyril> +1
Pierre: That's my recommendation. Merge the pull request and proceed.
Nigel: Any other views?
Gary: Note that large scale in WCAG refers to 18pt or 14pt bold or larger and the equivalent
… CJK fonts.
Nigel: Thank you
SUMMARY: We believe that SC 1.4.3 applies to the IMSC use case and requires a contrast ratio that exceeds that of 1.4.11, so we should note that is the primary WCAG guideline that applies to both text and image profile.
Nigel: I think if @gzimmermann still disagrees then we should arrange a call to understand his viewpoint better.
Pierre: I completely agree.
Address A11Y comments related to WCAG 2.1 imsc#526
github: https://github.com/w3c/imsc/pull/526
Pierre: Nigel, do you still want to continue the thread on this pull request?
Nigel: Yes, no reflection on John but I don't like this text at all.
Pierre: There were many hours spent on this text so my suggestion is to leave it alone,
… especially at the 11th hour, and move on. If you want to propose a replacement text,
… we can deal with it, but my recommendation is not to touch it.
Nigel: I do feel quite strongly, so I may propose an alternative.
SUMMARY: Action with @nigelmegitt to consider alternative wordings and whether or not to propose them.
Pierre: If you are going to do that, please do it immediately so we can close #526 or
… open a different ticket.
Nigel: OK, fair enough.
… I think I made the last comments on this PR so I was hoping for a response.
Pierre: My response was I would not change it, personally.
TTML2 Implementation Report
Cyril: I did a pass on the IR to add more rows but I think I need to do a second pass.
… I'm not sure the repo is up to date. For example I found files marked for 2nd Ed that
… are no longer in the test.json file. I wonder if those files should be mentioned there or not.
… I will check and let you know.
Nigel: Thank you.
… On the ttml2-tests pull requests, they are now all merged, so all the tests should be present.
… I see that you raised #238 Cyril to say some tests are in the manifest but not in the repo,
… and now you're saying there are also the converse!
Cyril: Yes.
… One other thing is we now have invalid presentation tests.
Nigel: Yes that's right I think, where we've defined fallback presentation requirements
… in the case of invalid data.
Cyril: So we need to add a 4th section.
Nigel: Yes that makes sense.
Cyril: Once I've made the second pass to the report it will be good if people could add
… their implementations. I am a little worried that we may not have enough implementations
… to meet the exit criteria.
Nigel: Yes, agreed.
… It will be easier to see that state of that until we have the report.
Meeting close
Nigel: Thanks everyone, we're out of time and have covered what we can for the agenda
… today. Next week I have added an agenda item already for an issue on the TTML Profile
… Registry, to add the RTP profile. It'd be good if we could tackle the long-outstanding
… issue 71 on that too, so I'd appreciate anyone's thoughts on how we can do that.
… [adjourns meeting]