xml-dev - Re: Subsetting/ Canonical Parsers/ XML Compliance/ etc.

[Rick JELLIFFE <ricko@geotempo.com>]

> Mike.Champion@SoftwareAG-USA.com wrote:

>      I'm having trouble reconciling this assertion with David
>      Brownell's results (see
>      http://www.xml.com/pub/2000/05/10/conformance/conformance.html
>      ).

When Dave put out that, I ran through the figures to put them into 
baskets and posted the results here on XML-DEV.   No-one gave any view
disputing my results, and I think they would only hold more now: that
conformance is generally excellent. (Even more so, now that MS has
improved.)

The test suite does not attempt to eliminate double counting.  So the
percentage figures are meaningless as indicators of quality.  What one
has to do is put them into baskets of functionality, to group them and
see where the problems occur.  

For example, the results showed that there was no real increase in
errors between validating parsers and non-validating parsers: the extra
features did not increase complexity to the point where it the features
could not be sustained.   

What was very noticeable from the early and crappy way that many
applications did not support xml:lang properly, was that developers did
not test their products systematically against test suites: that is
always going to be the nature of open-source code from independent
developers who probably have a particular application in mind they want
to solve.  It would not be solved by reducing the spec: if you just have
elements, data and comments someone would immediately try to get rid of
comments. 

Cheers
Rick Jelliffe