[relax-ng] Minutes RELAX NG TC telecon 2002-09-19

[Michael Fitzgerald <mike@wyeast.net>]

Minutes for a RELAX NG TC teleconference held 19 September 2002 at 9:00 am
EDT

The next telecon is scheduled for 3 October 2002.

Attending

David
James
John
Josh
Kohsuke
Mike

Not attending with regret

Makoto -- Congratulations on the birth of your son
Norm

Agenda

Type assignment

James: What do people think of making an association with elements and
attributes from a schema in an instance...rather than one QName? Make the
element or attribute arbitrary. The processor matches requested
annotations...
John: People could use xsi:type as well as any other arbitrary element or
attribute
Kohsuke: It's not as simple as that...collecting elements and attributes is
different
James: I agree, it's not exactly the same
Mike: I like this...we are talking about processor behavior with an element
or attribute input of the user's choosing, not just one attribute [such as
xsi:type]
James: The application decides what to do
Kohsuke: xsi:type can report element as well as attribute...not that useful
to RelaxNGCC or Relaxer
James: I'm not convinced, these are working at one level, they are meta
tools...this solution is not a panacea
John: Not that anything else is a panacea either
Kohsuke: So are we talking about keeping an element or attribute in memory
while we do processing...
James: I think it should be a runtime option. What you can express in an
attribute is...
John: ...limited...
James: limited...
John: Plenty of time, the element type is all you need
Kohsuke: But simplification throws away foreign elements or attributes
Mike: As the first step...
James: Yes as the first or second step
Kohsuke: This seems asymmetric to me
James: I think we are at cross-purposes; we want to do both attributes and
elements...the processor attaches following foreign elements as
annotations...are people in agreement that this is the way to do things
[associate arbitrary elements and  attributes in the schema and instance as
a way of doing type assignment]
All agreed

Business rules assembly

David: Both XSD and RELAX fail at this, assembling documents based on given
business rules
John: You can get this to work with a few rules, but when you have 27,000
rules, you can't figure out what goes with what...this is a place where
Schematron works...
David: Yes

Regex

John: I am going submit a new proposal on regex where it is an annotation
and the data pattern is given in a string; if a processor doesn't understand
the string, it is passed over as a simple string

Processing option

John: For Jing or any other processor, I suggest something like the opposite
of -f [feasibly valid wraps element, attribute, data, list in optional],
where you wrap any pattern in an interleave...
James: Use case?
John: This is a case for overvalid as opposed to undervalid