[
Lists Home |
Date Index |
Thread Index
]
On Thu, 16 Dec 2004 11:02:28 -0600, Peter Hunsberger
<peter.hunsberger@gmail.com> wrote:
> On Thu, 16 Dec 2004 16:17:46 +0000, tpassin@comcast.net
> <tpassin@comcast.net> wrote:
> > Roge Costello wrote -
> >
<snip>preamble</snip>
> >
> > > With the Bag approach properties are dynamically composed. The
> > > same properties can be associated with a different type. Further, the
> > > properties can be associated with multiple types.
> > >
> >
> > You can do the same thing with <person>...</person>, there is no essential
> > difference. It is just syntax, and it is a little more complex and harder to read.
> > It would be better to get the modeling straight about roles and relationships
> > with attributes.
>
> With a complex system I think your correct: everything comes down to
> modelling.
<snip>syntax digression</snip>
I need to make this a little clearer: syntax doesn't matter __if
you're careful__. Problem is, if you add in an XML Schema too early
in the process then syntax matters. The fact of the matter is, we
use an internal instance model that has concrete element names.
However, we don't force any schema on this. The trick here is to not
create a Schema and to match everything on type, not element name. At
that point:
<object type="address">
and
<address type="object">
are obviously interchangeable, (as is the RDF version). As I've said
before: I'll write no Schema before it's time...
--
Peter Hunsberger
|
