Re: [xml-dev] Re: determining ID-ness in XML

[Rob Lugt <roblugt@elcel.com>]

John Cowan wrote:
>
> > But having an implicit declaration would create two classes of Validity.
> > Existing (non xmlid-conformant parsers) would declare the document
invalid
> > whereas updated applications would declare it valid.
>
>
> Validity has been redefined before.  For example, it used to be valid
> to have two different attributes of type NOTATION on a single element,
> but this is now invalid in the name of SGML compatibility.  (It was
> always *meant* to be invalid, but formal invalidity entered by way of
> an erratum.)

True.  Another example was making duplicate enumeration values within the
same attribute invalid [1].  But all of these examples relate to practices
that were unlikely to occur in real life ~ therefore the impact was
negligible.

So, adding a new xmlid attribute has downsides either way:- you either need
to update your DTDs or you need to update your xml processors.  This is true
even for those people and applications that have no interest in being able
to address ID-less elements.

One advantage of the PI solution is that it has no impact on validity, and
is therefore opaque to existing applications and DTDs.

Regards
~Rob

[1] Errata E2 described in http://www.w3.org/XML/xml-V10-2e-errata