[tei-council] [TEI-DIR-WG] Outcome of TEI Council discussion on Text Directionality

[Martin Holmes]

Hi everyone,

This message is going both to the TEI Council and to the Text 
Directionality Working Group.

Following the discussion below in April and my presentation at the TEI 
Council meeting, nothing has happened partly because I've been very 
busy, but partly because I've been waiting for changes in the CSS 
Writing Modes specification to settle down a bit, and for an answer to 
the question of why bottom-to-top writing modes are not included in CSS 
Writing Modes. I haven't had any luck with the latter so far, although 
I'm waiting for a message to the W3C Style mailing list on the topic to 
be approved.

Meanwhile, I've been revisiting the proposal for new @rotate* 
attributes, and I wonder if there might be a better solution in CSS 
Transforms:

<http://www.w3.org/TR/css3-transforms/>

There are a number of reasons why I think this is a better approach:

1. It's an existing standard. Using an existing standard is always 
better than rolling your own, especially when we have already bought 
into CSS in the directionality part of the equation.

2. It's much richer than what we could achieve with our rotation 
attributes; in addition to simple 2D and 3D rotation, it allows (for 
instance) translation (moving an element relative to its base position), 
skewing, rotation around different origin points, and many other useful 
features.

3. It doesn't involve creation of any new attributes in TEI, so there's 
no need for argument about what elements should carry them.

4. User agents already have some support for 2D transforms and that 
support will increase over time. Where documents are being rendered in 
web browsers, it will be simpler for rendering pipelines to pass through 
CSS code where appropriate than to generate it based on TEI attributes. 
(This is a pragmatic point rather than a point of principle, of course.)

If we do recommend an approach to layout description based on CSS 
Transforms rather than on new TEI attributes, we'll be presenting a 
unified approach to the two separate but related problems (true text 
directionality and rotational features).

I'd be grateful to get feedback on this. If there's general approval for 
the idea, I'll update the workgroup Wiki page and the ticket:

<http://wiki.tei-c.org/index.php/Text_Directionality_Workgroup>
<https://sourceforge.net/p/tei/feature-requests/342/>

Cheers,
Martin

On 13-04-26 01:17 PM, Martin Holmes wrote:
> Doh!
>
> SHOULD BE:
>
> The existing one is for rotate-z, so we could say that where there is a
> single value, it applies to the z axis, and where there are multiple
> values, they are x, y, z (two values would presumably not be allowed).
>
> Cheers,
> Martin
>
> On 13-04-26 12:48 PM, Martin Holmes wrote:
>> HI Marcus,
>>
>> I'm replying to this through the mailing list, so that it gets archived
>> appropriately.
>>
>> I really like your idea of a single @rotate (rather than @rotation) with
>> three values, especially because it holds out the promise of
>> backward-compatibility with the existing @rotate attribute. The existing
>> one is for rotate-x, so we could say that where there is a single value,
>> it applies to the x axis, and where there are multiple values, they are
>> x, y, z (two values would presumably not be allowed). This would be even
>> more perfect if the current @rotate were x (the first of the three), but
>> I guess we can't have everything.
>>
>> On the other hand, if we create a new @rotation attribute, we can do
>> what we like with it, and perhaps deprecate the old @rotate after a
>> specified length of time.
>>
>> Regarding bottom-to-top, I appreciate the simplicity of going with pure
>> CSS for the moment, but I would like to find out why it's not in there,
>> and if there are plans to add it. If there's a principled objection to
>> it, we might well want to suggest alternatives.
>>
>> Cheers,
>> Martin
>>
>> On 13-04-26 11:51 AM, Marcus Bingenheimer wrote:
>>> Dear Martin,
>>>
>>> Thanks for that clear presentation and summary. I think we can be
>>> satisfied
>>> with the outcome.
>>>
>>> 1. As for the bottom-to-top issue I want to argue the following:
>>> CSS is likely to include bottom-to-top writing at one point, if not for
>>> Batak, Hanuno'o and Ancient Berber, then for the sheer ornamental
>>> purposes.
>>> This might come in tandem with a solution for writing along paths /
>>> Bézier curves.
>>> Instead of adding complexity to our recommendation we might be better
>>> off
>>> admitting incompleteness. In case someone is working on one of the
>>> extremely rare bottom-top scripts in TEI, the writing direction and its
>>> implementation is going to be a major issue in any case and has to be
>>> addressed specifically, probably somewhere in the header. I suggest
>>> we do
>>> not offer additional functionality for bottom-top, which might become
>>> obsolete as CSS evolves, but rather say @style should follow CSS to its
>>> present capabilities and recommend to record information on writing
>>> systems
>>> and direction that goes beyond that in the header for now.
>>>
>>>
>>> 2. Regarding the scoping of the @rotation-x-y-z:
>>> I am for the most "liberal" scoping as I want to rotate not only inline
>>> text but also images.
>>> By the way, have we considered to have only one @rotation which takes
>>> three
>>> values (x, y, z)? Something like @rotation="0 180 0" (180 along
>>> y), @rotation="50 0 90" (50 along x and 90 along z)
>>>
>>> all the best
>>>
>>> marcus
>>>
>>> On Thu, Apr 25, 2013 at 12:26 PM, Martin Holmes <mholmes at uvic.ca> wrote:
>>>
>>>> Hi all,
>>>>
>>>> This is a joint message to the TEI Council list and to the Text
>>>> Directionality Working Group list.
>>>>
>>>> We had a TEI Council meeting in Providence a couple of weeks ago, and I
>>>> made a presentation on the work we've done so far on text
>>>> directionality.
>>>> The slides of the presentation are here:
>>>>
>>>> <http://wiki.tei-c.org/images/**4/48/Text_directionality.pdf<http://wiki.tei-c.org/images/4/48/Text_directionality.pdf>
>>>>
>>>>
>>>>>
>>>>
>>>> and the minutes from the Council meeting are here:
>>>>
>>>> <http://www.tei-c.org/**Activities/Council/Meetings/**
>>>> tcm54.xml#body.1_div.1_div.3_**div.1<http://www.tei-c.org/Activities/Council/Meetings/tcm54.xml#body.1_div.1_div.3_div.1>
>>>>
>>>>
>>>>>
>>>>
>>>> At the meeting, I made the following three proposals:
>>>>
>>>> 1. That we ignore Unicode for the moment, mainly because it's not
>>>> relevant
>>>> for us (UTR #20 Unicode in XML and other Markup Languages
>>>> advises that bidi embedding (and presumably isolate) control characters
>>>> NOT be used in XML markup), and because it's moving VERY slowly on
>>>> vertical
>>>> orientation anyway (recent UTR #50 revisions have considerably
>>>> reduced its
>>>> scope in this regard).
>>>>
>>>> 2. That we formally adopt CSS Writing Modes, and provide examples of
>>>> how
>>>> to use it through the @style attribute.
>>>>
>>>> 3. That we create @rotate-x, @rotate-y and @rotate-z attributes to
>>>> capture
>>>> all manner of rotation (which can also help to handle edge-cases of
>>>> directionality such as boustrophedon).
>>>>
>>>> There was a generally favourable reaction to all three proposals, and I
>>>> was tasked with summarizing them to the Council list and to the working
>>>> group, which I'm now doing.
>>>>
>>>> These were some issues that Council would like to see addressed:
>>>>
>>>> 1. CSS WRITING MODES AND BOTTOM-TO-TOP SCRIPTS:
>>>>
>>>> It's notable that the CSS Writing Modes draft explicitly excludes
>>>> bottom-to-top vertical text: "Inherently bottom-to-top scripts are not
>>>> handled in this version. See [UTN22] for an explanation of relevant
>>>> issues." I personally don't see anything in UTN22 that justifies this
>>>> exclusion, and indeed very recent changes to CSS Writing Modes seem
>>>> to be
>>>> designed to fudge some accommodation for bottom-to-top into the system:
>>>> "The ‘sideways-left’, ‘sideways-right’, and ‘sideways’ values of
>>>> ‘text-orientation’ are provided for decorative layout effects and to
>>>> work
>>>> around limitations in CSS support for bottom-to-top scripts."
>>>>
>>>> Council asked me to contact the W3C working group and clarify the
>>>> exclusion of bottom-to-top scripts. Before I do that, I've been
>>>> trying to
>>>> figure out if there's any existing discussion of it on their public
>>>> discussion list:
>>>> <http://lists.w3.org/Archives/**Public/www-style/<http://lists.w3.org/Archives/Public/www-style/>>.
>>>>
>>>>
>>>> I'd appreciate any help with this that you have time to give.
>>>>
>>>> If the specification is leaving out bottom-to-top because of the
>>>> lack of
>>>> sufficient scripts in Unicode that are oriented that way, then maybe
>>>> Debbie
>>>> can let us know whether there are more bottom-to-top scripts on the
>>>> road
>>>> for inclusion in Unicode; we might be able to use that as an argument
>>>> for
>>>> more robust support for it in CSS Writing Modes.
>>>>
>>>> 2. PROPOSAL #3: WHAT ELEMENTS SHOULD HAVE THESE NEW ATTRIBUTES
>>>>
>>>> Currently, the TEI @rotate attribute (which really means rotate-z,
>>>> rotation around the z axis) is available only on <zone>. The three
>>>> attributes we propose, one of which would replace it, would be
>>>> provided as
>>>> a class. The question is what elements should be members of that class.
>>>>
>>>>   - The most conservative approach would be to keep it only to
>>>> <zone>, so
>>>> you could only use rotation if you're using the Facsimile module.
>>>>
>>>>   - Another would be to say that rotational features are inherently
>>>> topographical, and therefore all suitable elements in the genetic
>>>> editing
>>>> set (<line> etc.) might also bear them.
>>>>
>>>>   - Most liberally, we might say that these rotational attributes
>>>> may be
>>>> essential in any kind of transcription, so they should be available
>>>> on any
>>>> transcription-bearing element in <text> or <sourceDoc>.
>>>>
>>>> I'd be grateful for your feedback on this, and any insights anyone can
>>>> glean from the W3C style list archives regarding bottom-to-top text.
>>>>
>>>> Cheers,
>>>> Martin
>>>>
>>>>
>>>>
>>>>
>>>> --
>>>> Martin Holmes
>>>> University of Victoria Humanities Computing and Media Centre
>>>> (mholmes at uvic.ca)
>>>>
>>>
>>>
>>>
>>
>

-- 
Martin Holmes
University of Victoria Humanities Computing and Media Centre
(mholmes at uvic.ca)