More XML Parsing: Is anyone using CVS xml.el namespace processing?

[Mark A. Hershberger]

Recently I had a short conversation with James Clark about the
structures xml.el produces.

I asked him what he thought about the current (CVS) namespace-aware
processing.  Based on his feedback, I plan to submit changes that
will return an incompatible structure to the one currently in CVS.

Currently, when xml.el encounters a bit of XML like:

    <ns:xml xmlns:ns="uri:namespace" ns:attr="value"/>

it produces:

    (({uri:namespace}xml 
     (({http://www\.w3\.org/2000/xmlns/}ns . "uri:namespace") 
      ({uri:namespace}attr . "value"))))

At the time that I wrote this, I saw some W3 docs where this style was
used and copied it.  Some people here asked me why I did this instead
of something like (uri:namespace . "xml"), but I forged ahead.

Now, after my conversation with Mr. Clark, I've been persuaded that I
was wrong.  At his suggestion, I'd like to change the above xml
representation produce the following:

    (((uri:namespace . "xml")
     ((((http://www\.w3\.org/2000/xmlns/ . "ns") . "uri:namespace") 
      ((uri:namespace . "attr") . "value")))))

As Mr. Clark said:

    ... there are typically not very many different namespace URIs, so
    keeping them in Emacs symbol table is not a problem; in the
    returned representation of the XML, the namespaces would be
    shared, but strings are mutable in Emacs, which is kind of ugly.

Where there is no namespace given:

    <xml attr="value">

It would produce the following:

    (("xml"
      (("attr" . "value"))))

Unless there are major objections, I'd like to repent of my previous
code and submit changes to produce the above.

Mark.

-- 
If you want to know who is funding terrorists, look in the vanity
mirror as you turn the key of your SUV.
    -- http://philip.greenspun.com/politics/israel/