- From: Daniel Veillard <daniel@veillard.com>
- Date: Mon, 16 Jun 2003 11:48:38 +0200
- To: Elliotte Rusty Harold <elharo@metalab.unc.edu>
- Cc: daniel@veillard.com, www-xml-xinclude-comments@w3.org, "sandra.martinez" <sandra.martinez@nist.gov>
On Sat, Jun 14, 2003 at 09:53:28AM -0400, Elliotte Rusty Harold wrote: > Short of that, if we're going to have sensible results, then there > needs to be an assumption that the conversion from the infoset to the > actual XML documents presented as test results should not add > information items to the output that were not present in the input > document. Since the Document Type Declaration is an Information Item, > I think these test results are actively wrong. They strongly suggest Yeah, it's a bug, I agree. > that the expected output contains an information item it does not in > fact contain. > > The problems much trickier for the cases where it's necessary to add > a document type declaration in order to represent the notations, > unparsed entities, attribute types and other infoset augmentations > performed by the DTD. In that case, you may well be dealing with a Another problem is that the xml:base which needs to be added is dependant on the location of the test-suite at runtime unless one proceed to do some convoluted URI relativization of the xml:base values. > genuinely unserializable infoset; e.g. one that has a notations > property on the document information item but not document type > information item. In hindsight, it was probably a mistake to make the > notations and unparsed entities properties of the document > information item rather than the document type information item, but > we're stuck with it now. I don't think this can be fixed short of > going to a more complex test framework like that used for XSLT > conformance testing. Well in general, I think the goal is more to exercize the spec being tested and have unanbiguous results w.r.t. that given specification, chasing a more general goal which is to provide a regression tests valid for all implementations is far more costly, I'm not sure the Working Group really have the cycles to reach that goal, nor that it's really what we want to guarantee. Daniel -- Daniel Veillard | libxml Gnome XML XSLT toolkit http://xmlsoft.org/ daniel@veillard.com | Rpmfind RPM search engine http://rpmfind.net/ http://veillard.com/ |
Received on Monday, 16 June 2003 05:48:47 UTC