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

["Jonathan Borden" <jborden@mediaone.net>]

Simon St.Laurent wrote:

> >> Maybe that should be fixed in XSLT, not in XML?
> >>
> > It would be easier to fix it in XSLT if it was first fixed in the
InfoSet...
>
> Ouch.
>
> Maybe it's time to abandon any notion of type specification, and just
> stick to the small bit of information that can be gleaned from a
> document instance minus DOCTYPE, even if it rules out certain
> possibilities.
>

Being too quick to discard DTDs is the perfect example of the premature
ejection of a core part of XML 1.0, at the same time piling on new
'features' for XML. The whole point is that one can _already_ implement
in-line ID attribute declarations (using the internal DTD subset). As I
recall, we have _already_ gone though a long process in an attempt to
replace DTDs with some type of XML syntax, yet today we still are having
these discussions. The ability to define an identifier unique within a
document seems like a straightforward and reasonable requirement. Do XML
Schema or RELAXNG solve this apparently simple problem in an equivalent
fashion to the DTD?

I think that XML 1.0 _taken as a whole_ is far better designed than most
people realize, and what some people see as a feature to be discarded with
their left hand, they seem to ask for in another form with their right hand.
It is time to better use what we already _have_ which should get around the
global problem of too many XML specifications that sort of but don't
completely work together.

Jonathan