Re: [xml-dev] My report on experiments with unused namespaces

[Michael Kay <mike@saxonica.com>]

> It's only a concern if you choose to be concerned about it. As all 
> commenters have noted, there is no definition of "unused". Any markup 
> in the document may be used by someone.

There are several reasonable starting points for defining what parts of 
an XML document are information-bearing: two that agree reasonably 
closely (though not 100%) are the XML canonicalization spec and the XDM 
data model. These both provide answers to questions such as whether 
CDATA boundaries are significant (no), whether whitespace within a start 
tag is significant (no), whether the choice of hex or decimal character 
references is significant (no), whether whitespace in 
element-only-content is significant (yes, with caveats), and whether 
namespace prefixes are significant (yes, with caveats). As a first cut, 
information has to be treated as significant if it survives XML 
canonicalization. Which means it MAY be used by someone.

A stricter view might be whether it SHOULD be used by anyone; one might 
imagine a stronger form of canonicalization that, for example, moves all 
namespace declarations to the outermost element, changes the prefixes, 
and deletes namespace declarations that are not used in any element or 
attribute names. You might take the view that people should write their 
applications in such a way that they continue to work after such a 
change; and others might legitimately take a different view. This is in 
the realm of "best practice", where no two people will agree.

Michael Kay
Saxonica