Re: text-wrap balance

One of the problems I foresee with defining balance justification is that
the algorithm used for balancing is implementation dependent [1]:

The exact algorithm is UA-defined.


This puts balance into the same category as line breaking in general, which
is not testable unless a specific algorithm is applied. In TTML, we defined
[2] a (semantics only) feature designation, #lineBreak-uax14, which
requires use of UAX14 [3], thus permitting testing and a degree of
interoperability; however, the only published formal algorithm for balance
that I'm aware of is contained in Knuth's *TeX: The Program* [4]. However,
I did find a poly-fill for balancing [5], from which one might be able to
generate a formal algorithm description. I rather doubt, however, that CSS
will define a normative algorithm (like UAX14) that could be referenced by
TTML or VTT.

G.

[1] https://www.w3.org/TR/css-text-4/#text-wrap
[2] https://www.w3.org/TR/2018/REC-ttml2-20181108/#semantics-line-breaking
[3] http://www.unicode.org/reports/tr14/tr14-17.html
[4] https://www.amazon.com/Computers-Typesetting-B-TeX-Program/dp/0201134373
[5] https://github.com/adobe/balance-text

On Sat, Jun 15, 2019 at 7:27 AM Silvia Pfeiffer <silviapfeiffer1@gmail.com>
wrote:

> The use case for text-wrap: balance is particularly nice for captions
> where best practice in caption authoring requires the author to balence the
> rendered lines for better readability.
>
> I wonder if this would also be relevant to TTML? Would it make sense to
> lobby the CSS group for it?
>
> Kind regards,
> Silvia.
>
>
> On Sat., 15 Jun. 2019, 3:39 am Philippe Le Hégaret, <plh@w3.org> wrote:
>
>> Forwarding here with permission:
>>
>> -------- Forwarded Message --------
>> Subject: Re: Fwd: Re: text-wrap balance
>> Date: Fri, 14 Jun 2019 09:45:37 -0400
>> From: Chris Lilley <chris@w3.org>
>> To: Philippe Le Hégaret <plh@w3.org>, Fuqiao Xue <xfq@w3.org>
>>
>>
>> On 2019-06-13 16:55, Philippe Le Hégaret wrote:
>>  > Chris, Fuqiao,
>>  >
>>  > the Timed Text Group/WebVTT is wondering what to do with text-wrap:
>>  > balance. Do you know or can you find the story behind it? WebVTT
>>  > relies on that value but if no one implements it, there isn't much
>>  > point...
>>
>>
>> text-wrap: balance (and no-wrap) was proposed for CSS by Adobe. There
>> used to be a proposal on their site [1] but that has disappeared.
>>
>> It is not in CSS Text 3 but was added to CSS Text 4 [2][3] and Adobe
>> also maintains a JQuery plugin [4] which implements it. There are no wpt
>> tests for the text-wrap property [5]
>>
>> I have seen other implementations of line-balancing in JS. Other plugins
>> or polyfils will be easier once Houdini provides the ability to measure
>> the length of a line.
>>
>> There are a couple of open issues: [6][7][8][9]. From [7], Apple seems
>> to be slightly against due to the iterative algorithm (number of passes
>> is unknown, and interaction with text fragmentation is unclear).
>>
>> Current status seems to be a bunch of web developer interest, no
>> implementer interest. The spec might be improved by a couple of good,
>> visual examples. Needs evangelism to demonstrate need, I suspect.
>>
>> [1] https://adobe-webplatform.github.io/balance-text/proposal/index.html
>> [2] https://www.w3.org/TR/css-text-4/#text-wrap Sept 2018
>> [3] https://drafts.csswg.org/css-text-4/#text-wrap
>> [4] https://github.com/adobe/balance-text
>> [5] https://github.com/web-platform-tests/wpt/tree/master/css/css-text
>> [6] https://github.com/w3c/csswg-drafts/issues/3047 (from frivoal
>> <https://github.com/frivoal>)
>> [7] https://github.com/w3c/csswg-drafts/issues/2528 (from tobireif)
>> <https://github.com/tobireif>
>> [8] https://github.com/w3c/csswg-drafts/issues/1975 (from palemieux
>> <https://github.com/palemieux>)
>> [9] https://github.com/w3c/csswg-drafts/issues/803 (from frivoal
>> <https://github.com/frivoal>)
>>
>>
>> On 6/13/2019 4:33 PM, Silvia Pfeiffer wrote:
>> > FWIW, I think asking a status update from the CSS group on this
>> particular
>> > feature would be great.
>> >
>> > Cheers,
>> > Silvia.
>> >
>> >
>> > On Thu., 13 Jun. 2019, 9:09 pm Philippe Le Hégaret, <plh@w3.org> wrote:
>> >
>> >>
>> >>
>> >> On 6/12/2019 10:43 PM, Pierre-Anthony Lemieux wrote:
>> >>> Hi Gary et al.,
>> >>>
>> >>> In recent publications the group has gone to great lengths to make
>> >>> sure that at least two implementations passed each test, whether for
>> >>> exotic or trivial features.
>> >>   >> Why would it be different here? Have the criteria changed? Should
>> >>> future versions of TTML and WebVTT have to meet a lower threshold of
>> >>> "proof-of-concept"?
>> >>>
>> >>> I think the group needs to be consistent, one way or another.
>> >>
>> >> I thought the exit criteria was 2 implementations of each feature.
>> There
>> >> is a difference between tests and features.
>> >>
>> >> However, I'm not suggesting that the difference matters in this
>> >> particular test, given that the spec is clear on this particular test.
>> >> If the value balance isn't supported today, what are the
>> implementations
>> >> doing instead of balance? Chromium doesn't seem to have a bug report on
>> >> balance for example. Should we ask the CSS Working Group on the status
>> >> and stability of balance? (Gary, I'm happy to go on fact findings if
>> >> you'd like)
>> >>
>> >> Philippe
>> >>
>> >>
>> >
>>
>