Simple XML conformance

[David Brownell]

Peter Murray-Rust wrote:
> 
> Is there a definitive resource anywhere which explicitly states what
> behaviour can be expected from various types of parsers? I know it is
> inferable from the spec, 

There's a test matrix up at the front of the OASIS/NIST test report
that covers some of this, but most if it must be gotten from the spec.
That test report is big (571K) and an HTML rendition is at

    http://www.oasis-open.org/committees/xmltest/xmlconf.htm

If you don't want to load 571K, then just load sections 1 and 2; the
test matrix is in section 2.

(The source is an XML document ... you might find it interesting to
see how many of your parsers handle it correctly!  :-)


Luckly several of your examples don't call for "inference" from the
specification -- when the spec says something is optional, it means
it, so parsers objecting to missing XML declarations or (specifically)
DTDs are clearly in error.  (If there's no DTD, a validating parser
will certainly find an error in a document -- but it's not that the
DTD is missing, it's all the undeclared names in the document!)

That parser that interpreted <!-- %pe-refs-in-comments; --> really
would benefit from reading the recently published errata ... it fixed
some significant omissions from the original XML spec in that area.
Any parser implemented from just the spec would have that problem.
(I got suspicious when my parser couldn't parse the XML spec itself!) 


>	but I suspect that not all implementers have taken
> identical interpretations. I would ideally like to have a matrix of parsers
> against standard "correct" [not always "valid"] documents and see how many
> conform.

Again, look at that test matrix.  The whole OASIS/NIST suite basically
took a thousand or so documents and slotted them into the boxes in that
matrix ... the process of testing is mostly taking a parser and seeing
how it compares with the expected results.

- Dave

xml-dev: A list for W3C XML Developers. To post, mailto:xml-dev at ic.ac.uk
Archived as: http://www.lists.ic.ac.uk/hypermail/xml-dev/ or CD-ROM/ISBN 981-02-3594-1
Please note: New list subscriptions now closed in preparation for transfer to OASIS.