Re: using namespaces to version

[Warren Hedley <w.hedley@auckland.ac.nz>]

Hi folks,

First of all, thanks to everyone who's participated in this thread so far.

If I can direct attention back to my original problem. For my particular
project (CellML - http://www.cellml.org/ , if anyone's interested), the
versioning process will almost definitely affect the validity of documents
(i.e., if there was a schema, it would change), and the expected behaviour
of processing applications.

Backwards compatibility is likely to be on an element-by-element basis. For
instance, the content model and semantics of the <connection> element will
probably not change in the near future, but the <variable> element almost
definitely will be extended (although the CellML 1.0 way of doing things
will probably not be made invalid). Because this element affects the
definition of a mathematical model, it is highly unlikely that a CellML 1.0
processor will be able to process the extended form of the CellML 2.0
<variable> element in a manner conformant to the CellML 2.0 specification.

Some of the validity constraints on CellML are currently too complex to be
expressed using DTDs or schema (AFAIK). Instead we have a technical
specification which describes the validity constraints and semantics for
all of the elements (similar to MathML's specification). For these reasons,
I expect the namespace URI will point to human-readable documentation
instead of a schema or RDDL document.

Other than the XPath problem that David Carlisle raised (which is easily
solved), I can still not see any good *technical* reasons why I shouldn't
use namespaces for versioning. In fact, it's quite attractive because
CellML is likely to be available and processed in fragments, in which there
won't use a DOCTYPE declaration, and I don't want to define some kind of
version attribute on every language element.

Comments, anyone?

-- 
Warren Hedley