Re: Something like xml:dir (was: Re: [css3-writing-modes] direction vs. writing mode in markup vs. style)

On Mon, Oct 25, 2010 at 10:43 PM, "Martin J. Dürst"
<duerst@it.aoyama.ac.jp> wrote:
> On 2010/10/26 2:54, Tab Atkins Jr. wrote:
>>
>> On Mon, Oct 25, 2010 at 5:59 AM, John Cowan<cowan@ccil.org>  wrote:
>
>>> Hmm.  Do you want the XML Core WG to take up the idea?  We own the
>>> xml: namespace, and we aren't that busy these days.
>
> I wouldn't want that people talk about something like xml:dir as "not
> exactly a good idea, but the XML WG wasn't too busy, so they cooked that up,
> too". Please think about it on the merits. Also, please think carefully
> about what to do to make sure that designers of XML vocabularies will use it
> in the places it makes sense, and won't use it in the places that don't make
> sense.
>
>> That actually sounds good in general, though see the discussion in the
>> HTMLWG surrounding the changes suggested by the "Additional
>> Requirements for bidi in HTML" group.
>
> Yes, I'd definitely wait for that discussion to settle down (or participate
> if you are interested). In addition, please carefully analyze what an XML
> vocabulary that is very data-oriented (as opposed to document-oriented)
> might need. The original (ca. 1995) design for bidi in HTML was aimed only
> at static documents; the current additional efforts try to cover more
> dynamic documents and therefore cover data in one way or another, and that
> may or may not be enough.

Right; fantasai and I were actually part of that group, so we
definitely participated.  ^_^


>> (This does bring up potentially painful issues similar to @lang vs
>> xml:lang, or charset handling, but if it's handled in a similar way it
>> will at least be usable.)
>
> I hope it will not be handled in a similar way. The problem with xml:lang in
> (X)HTML was that people assumed because there was an xml:lang, that needed
> to be used, and they also worried about interpretation of the same document
> as HTML, so lang was also required. The right way is to use something like
> xml:dir for new XML vocabularies, and let HTML and XHTML stay with the
> current unprefixed attributes. All other XML vocabularies that already have
> some bidi mechanisms should also stay with the attributes (or elements) they
> already have.
>
> Another concern is that for bidi attributes in HTML, there is a default
> stylesheet. So would we need a default stylesheet for XML to cover something
> like xml:dir?

No; bidi rendering *should* occur solely in the rendering engine, and
shouldn't be exposed to CSS at all; this is not a styling issue any
more than "should I interpret this text as unicode or ascii?" is.  The
CSS support for the bidi attributes was solely so that XML languages
could define their own bidi attributes and get them to respond
correctly in a CSS-based processor.

If XML supports directionality annotations natively, then the use-case
for CSS directional properties disappears.  We of course can't get rid
of the properties now, but we can obsolete them, and avoid carrying
them into future drafts (keeping them as just a CSS2.1 quirk) or
expanding their functionality.

~TJ

Received on Tuesday, 26 October 2010 16:48:19 UTC