relative URIs and local lexical scopes: "Unique Base"?

I was trying to explain the issue to someone recently, and came
across a different analysis. The issue is that URIs are mainly
dynamically scoped: the relative URI is interpreted against the
base at the time when the base is known to be valid.

Namespace names, however, establish a global lexical scope to
allow combining and linking the local lexical scopes of different
XML fragments and components; relative namespace names seem to be
trying to establish locally unique lexical scopes. The problems
are exacerbated because of XML macro capabilities in transformation
languages and entity references. The difficulty of creating local
lexical scopes for macro processing was one of the hard problems of
the 80s in language theory; for example, it was the subject of Kohlbecker's
PhD thesis.

ftp://ftp.cs.indiana.edu/pub/scheme-repository/doc/misc/macros-03.txt


Proposal:  Unique Base

   Relative URI references are allowed as namespace names. However,
in lieu of any other XML mechanism for establishing a base URI for
namespace names, the base to be used is to be constructed explicitly
for the containing document instance as a completely unique URI base.
The algorithm for uniqueness need not be part of the specification,
but, for example, it could be based [sic] on the URL of the namespace
processor and the date and time, identity of the document instance
being processed, etc.

The idea is to insure that an
  <a xmlns="frob">
in one document will *never* match an
  <a xmlns="frob">
in another document, since the relative URIs in the two documents will
be guaranteed to have a different base.



  

a relative namespace name is to be resolved
-- 
http://larry.masinter.net

Received on Saturday, 24 June 2000 15:24:42 UTC