Re: [xml-dev] three-vowel element names (was Re: [xml-dev] xml:href,xml:

The point is that xlinks major (fatal really) failing is that it forces
attribute names on the vocabularies that use it. (Its bad enough that
xml:id forces a name but the issues are worse with link attributes)

I agree, although I think an AF-like approach to linking should also take on IDs, which are just anchors. xml:id would of course have to be the default, recognized way to specify IDs/anchors (I know some would say these are not the same thing, but I believe treating them as such is a pragmatic simplification).

architectural forms which pre-dated xlink from SGML had a much more
extensive vocabulary that allowed you to impose linking semantics on
_existing_ markup without the markup having to change.

So you want to be able to say things like

a/href is a normal link, img/@src is an embed link.

But the recognition patterns might need to be more advanced such as
Liam's somewhat extreme example.

I think the point is that if you have some flexibility in specifying what attributes have the semantics of link/@href or link/@rel, you might as well provide all the flexibility of at least XPath. Of course Liam's example would need more than XPath 1.0, and would also need regex facilities, but I say "why not?" to that (besides aprocryphally having created an additional problem ;) ).

FWIW I think such an AF-tool would be dead easy to implement in XSLT, and as I recall both Jeni Tennison and John Cowan have given it a go in the past. Those ideas were never really taken up, but I wish they had been. Maybe an AFng could be a nice addition to the MicroXML family of specs.