Re: [xml-dev] IDs considered harmful or why keys might be betterthanIDs...

[David Brownell <david-b@pacbell.net>]

> Right, but changing XML 1.0 is precisely what would be required, i.e. to
> require that all XML processors report attribute types, at least ID
> attribute types. If everyone's already doing it, then it's not that big of a
> deal, but it makes sense to include it in the spec, perhaps as an erratum.

It's been pointed out before that the XML 1.0 REC had some gaping
holes with respect to reporting requirements.

If you look at the XML source for the original REC you'll notice that for
some reason the section (3.5 it would have been) requiring processors to
report at least the element and attribute names (not attribute values, not
attribute types ...).  After several years to fix that simple problem, one must
concludes that it's rather unlikely ever to get fixed.  The Infoset at least
characterized the important info, even though it punted on firming up
reporting requirements.

On the other hand, fast'n'loose spec work at W3C isn't a novelty;
it's what triggered this debate.  All names begining with "XML" (in
any case combo) are reserved for future versions of that REC ... yet
we have "xmlns", "xml:base", and so on for "layered" specs.  "xml:id"
(etc) would be no more of a stretch.

- Dave