Lists Home |
Date Index |
Tim, you've been playing in this field much longer than I have. So
tell me where I got things wrong.
Tim Bray <firstname.lastname@example.org> writes:
> I always felt vaguely guilty
> about the existence of attributes, but that <a href="x">y</a> idiom
> seems so smooth and more idiomatic than any other syntax I can
> imagine, that it long ago reconciled me to them. I haven't been able
> to work out the abstractions and metaphysics of why this feels so
> right, beyond vague hand-waving
I'm not sure where I got it from, but I seem to have this notion that
character data are what a human reader should see, while markup,
including attributes, is something of a hint to software.
> But then why not five?
There are characters in a document and (meta)information about the
characters. Some of those metadata can be predefined, e.g.,
html:title, and that leads to element tags; sometimes you need free
text, such as for an href, which leads to attributes.
So I see three things - text, predefined metadata and free-form
metadata. I can't immediately think of three more kinds of metadata
I'd want in a document.
> I think unordered attributes, and dictionary type
> structures to model them in software, are way out on the plus side of
> the cost-benefit equation as a design decision in SGML and XML.