Re: xml:* attributes

Uche Ogbuji scripsit:

> Actually, I don't see it this way. I don't see avoiding a
> fragmentation of layers on top of μXML as a goal at all. John
> proposed the "self-contained" goal and I think I have an idea why,
> but I'd leave it to him to say for sure, but John has also spoken of
> Examplotron as well as MicroRNG (though more lately the former), which
> would count as a bit of "fragmentation" in itself.

Indeed.

By "self-contained" I meant mostly "self-contained with respect to
the core XML specs such as XML 1.0, XML NS, xml:id, xml:base, and the
Infoset".  I never meant to imply that everything would be either in
the MicroXML spec or else purely application-dependent.  I think that's
absurdly rigid.  (Should MicroXML also include MicroUnicode?  I think
not.)

> > In pure micor-xml, xml:id attributes should either work or should be
> > banned, and since no one appears to want to suggest that the data
> > model should include ID typing, that means banning.

Again, too rigid for me.  Why not say that "p" elements must work or be
banned?

Anyway, what counts as "working"?  The xml:id spec is careful not to
require that xml:id values be unique, though it does require that they
be NCNames.  (I was against that requirement myself, but it was pointed
out that upper layers might break if they got ids that were not NCNames;
still, in the absence of validation this is still possible).

Rather, the point of xml:id is that it is a semantic signal of the
intent of this particular attribute: it is used to identify its element.
That is independent of validation issues such as format and uniqueness,
which indeed can be handled by upper layers in various ways: reject
non-uniqueness, or use the first identical value as HTML and XPath do.

> > If another specification needs micro-xml + xml:id it should extend
> > the language and the data model at the same time.
>
> I think that is the *real* path to fragmentation, having each
> layer change syntax in a way incompatible with the layer below it
> (i.e. μXML bans this attributes but μXML-id-prime re-allows them).

+1

--
John Cowan  <cowan@ccil.org>  http://www.ccil.org/~cowan
        Raffiniert ist der Herrgott, aber boshaft ist er nicht.
                --Albert Einstein

Received on Thursday, 16 August 2012 16:39:14 UTC