Re: 2.1 a-d: Link Recognition by Reserved Attribute?

The question is whether to recognize link elements by attribute, GI,
or both:

Jon Bosak wrote:
>I'm not following the reasoning for having both forms.  If the
>reserved attribute method works, why do we need the reserved GI?  If
>the reserved GI method has advantages, why use the reserved attribute?

Well, if we're doing it by attribute, (in effect AF) then the AF is
going to have to have a name, so why not just allow its use as a GI?
I would think that the *easiest* way to do an XML link, both to explain
and to do, would be just to adopt a hard-wired element nicely predeclared
in the spec.  This would be my first choice.  I think would be desirable
if everything whose GI was XML-TLINK was an XML TLINK, and every XML TLINK
had a GI of XML-TLINK.  (or whatever we call 'em)

But in the XML discussions, there was some considerable reluctance to
predefine particular elements; and it seems the conventional wisdom that the 
*right* way to do these things is with architectures; for reasons of 
namespace infringement, and general good karma.  Thus we wrote this 
possibility into the spec.

Having written this, I'm beginning to discover a new hard-minimalist 
enclave I can gleefully retreat into; why not always do it via GI?  That
question is not rhetorical; perhaps someone can explain the downside to
this, and the upside to supporting AF-like attribute signalling?

 - Tim