Thanks,  I suspected something from the distant SGML land as it seems entrenched in the DTD history ..

But obviously any sane person would assume

                xml:id="D000001"   == xml:id="D1"

:) 1/2

On 29/04/2011 15:22, David Lee wrote:

First off I know asking "Why" for anything standards related is silly.

But given that, could anyone give me a rationale or history for restricting xml:id (or the ID type from DTD) to be NCName which then has to start with non-numeric ?

I seem to recall asking this once, and being told that the history lay in the SGML rules for abbreviated syntax. Something like allowing <e id="xyz"/> to be abbreviated as <e xyz>. Not that this in itself would make a numeric identifier ambiguous, but it would account for requiring an ID value to have the same syntax as an attribute name.

I guess another justification is that it prevents any debate about whether xml:id="01234" is a duplicate of xml:id="1234".

Michael Kay
Saxonica