Hi Dave,
On Jun 5, 2006, at 1:23 PM, Dave Pawson wrote:
>> Anyway, to get to my point on the caption proposal, I recognize that
>> adding an attribute link is a fairly painless solution, but I also
>> think it's not a good long-term one, not only because it will conflict
>> with the efforts of the metadata SC.
>
> How please Bruce?
A "describedBy" attribute is exactly the sort of ground that we intend
to cover with metadata. It associates an object with a metadata
description. In this case, that description is just plain text, but in
other cases, it could be full metadata records (identified by uri). In
fact, one of our use cases involves automatically generating caption
from embedded (probably RDF) metadata descriptions.
<http://wiki.oasis-open.org/office/Intellectual_Property>
>> My suggestion would instead be to use natural containment, and to add
>> a
>> caption element to ODF 1.2; e.g.
>>
>> <text:p>
>> <draw:frame> <!-- outer frame -->
>> <draw:text-box>
>> <draw:frame>
>> <draw:image>
>> <office:binary-data>picture data
>> here....</office:binary-data>
>> <text:caption>Caption Text</text:caption>
>> </draw:image>
>> </draw:frame>
>> </draw:text-box>
>> </draw:frame>
>> </text:p>
>
> That infers association. ours presents it as direct, stated.
> How do you see one as better than the other?
OK, let's break apart two separate issues: association and semantics.
On the second, my contention is that "caption" is an important semantic
structure, and so deserves its own element. To say something is
"describedBy" is -- per above -- rather vague. So notwithstanding the
question of how this lines up with a broader metadata effort, I think
we need a standard way to say what kind of description it is.
Note: there are other important semantic terms -- quote, blockquote,
etc. -- that come to mind that could also use some sort of
standardization, but for those I'd be content with simply controlled
style names. That'd be possible here too of course.
WRT to association, I think if we start down the path of using linking
to associate content with description, things could get really messy.
It's really a different design approach.
For example, I mentioned that I didn't like the way sectioning
currently works. Ideally, we'd have proper sectional structures a la:
<section>
<p>...</p>
</section>
The standard way to include heading information for those is
containment; e.g. in XHTML 2.0, IIRC it's:
<section>
<h>Heading</h>
<p>...</p>
</section>
... DocBook:
<section>
<info>
<title>Heading</title>
</info>
<para>...</para>
</section>
... etc.
Aside: the above could benefit accessibility, yes?
If we ARE going to go down the road of using attributes to make this
association, then I think:
a) we need to give it much more thought so that we adopt a consistent
approach to these problems
b) it needs to happen in conjunction with the metadata effort
Thoughts?
Bruce
