SV: Updated proposal outlined

I've inlined my comments below.

-----Ursprungligt meddelande-----
Från: Arnold, Curt [mailto:Curt.Arnold@hyprotech.com]
Skickat: den 25 maj 2001 20:55
Till: 'www-dom-ts@w3.org'
Kopia: 'xmlconf-developer@lists.sourceforge.net'
Ämne: Updated proposal outlined


A URI scheme for the existing staff* files and the NIST/OASIS XML v2
conformance suite would be helpful.  We should probably try to minimize the
desire of people to add new test documents to the DOM
test suite.  If there isn't coverage of the structural feature, then it
probably should go into the XML conformance suite.  Ideally, the URI's would
be resolved against a local copy in a JAR or other
archive.

One of the reasons that I preferred defining the tests in XML Schema and
then generating the DTD is that the schema can be used as a resource in the
transforms.  The transforms would only need to have
templates for the language elements (the <asserts>, <assigns>, etc).  All
the DOM methods could be generated by a generic template that checked the
schema for the "type" of the element to determine if
the element was a property accessor, property mutator or function and the
order of parameters.  Without obtaining that information from the schema,
that has to be replicated in the transforms.

*** No argument on Schemas being better. DTD are just an old habit with
better support in general among editors. In either case, I think we should
have both. Needless to say, Schemas will add functionality, where DTD will
at least be of low-level syntactical help to the test developers.

The problem with not explicitly declaring the variables comes on platforms
where there is not a common ancestor class (like Object in Java).  If I was
converting to C++, then it is really significant
that I can determine that an variable is a String, Node, integer or the
like.  Trying to guess the appropriate type really ups the complexity.  XML
Schema's <key> and <keyref> could be used to assert
that all var and object parameters were defined.  However, I haven't
explored if they are supported by the current schema aware editing tools or
whether they can be done in a manner that would be
legal both in Oct 2000 Schema (for compatibility with current tools) and the
recommendation.

On case submission, I'd prefer a system where any test submission becomes
immediately publicly visible.  Whether it is a good test, unneeded test, or
a just plain bad would be decided in the WG in due
time and expressed in RDF.

*** This list will be the principal forum for submitting tests, so we will
have 100% visibility.

If a test is hidden until the WG passes judgment on it, then there is not a
chance for others to suggest enhancements, raise issues, etc.  Then it has
to cycle the WG if there were any issues.

Since you want to basically give the public write access, CVS doesn't seem
appropriate since you'd have to give everyone too much power to do damage.
How about something like Manila or SharePoint?
Maybe just in the test collection process.

*** Sure, let's speak about this on the telcon. You have plenty of
experience with CVS to bring along, so I look forward to discussing it then.

Received on Friday, 25 May 2001 15:06:23 UTC