From: "Simon St.Laurent" <simonstl@s...>
 
> I'm also maintaining both an archive of these drafts and the currently
> open (unsubmitted) draft at:
> http://simonstl.com/ietf/

> Comments and suggestions are welcome.

Looks good. It still falls down for the same reason that the whole
"out-of-band signalling using MIME headers" falls down:
there needs to be some convention (other than by parsing
the whole document including subentities) for a webserver to 
construct the information in the first place. 

Is there value in constructing pipes when there are no
feasible connectors?   At the moment, the WWW basically
runs on 
  file extension -> MIME type -> type handler
We know that a document is xslt not because of its top-level
element but because it is .xsl or .xslt.

The obvious way to have access to namespace precis in
a document (saved as a file) is to bubble up all namespace
declarations to the top-level element.

But if we need to canonicalize our documents in this way
in order to have the namespace details in a convenient
place for a webserver to re-extract it, why not just
send the document like that in the first place, and let
the application decide how to handle it? 

It seems to me that "best practise" XML should include
having xmlns:* declarations for all namespaces used
in the document at the top level element, even if 
the namespaces are used by default declarations
within the document rather than the dummy prefixes
at the top-level.

Would it make processing and dispatching easier if
XML tools did this?

Cheers
Rick Jelliffe