Re: xmllink-970406 various

>The circumstance I'm wondering about most is this:  I download an
>XML instance, construct (a) XML TEI pointers into it using fragment IDs
>and (b) XML TEI pointers into it using queries.  How do I know that
>either will work in practice over the Net?  In the case of (a) it will suffice
>that the server doesn't grok the pointer but anyway sends me the entire 
>instance - that appears to work in a quick random test with HTML.  In the case 
>of (b), quick random testing reveals that some servers return the
>entire instance (not what was expected, some error recovery will
>be needed there), or fail to return anything but an error message.

Right, and this is the core of my disagreement with using queries.
We either standardise something, or nothing will be gauranteed to
work, in which case, it's probably better to stay silent for the

Content negotiation *may* help here, but is currently not useful
for detecting server capabilities like the above.