Re: schemaMapping issues breakdown (issue 6)

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Can we consider schema mappings at a very general level, where they convert to/from data items and concept instances?
If so, they can be placed at toplevel in the specification since they can be reused (it is needless to mention a mapping
every time a modelreference is used for an element having the same type), and not only close to the (most relevant?)
modelReference.

Also: (questioning extension attribute?)

I am really not fond with constraining tools to parse or derefer uris to select modelReferences if needed (as was agreed
when closing the issue 5 lately - I fear I failed to grasp all the implications of this vote), but I agree that a
significant amount of relevant information can be present in such uris (namespace, versions, etc...)

When it comes to mappings however, the uri will not be exploitable to tell apart which concepts and/or types it refers
to. Hence, dereferencing will always be required for tools to know which mappings they can use. If we wish to allow for
smooth evolution, we might consider to raise the status of mappings from extension attributes to model elements, so that
the concepts/types they apply to in the current spec can be precisely listed, without requiring to dereference the uri.

As for model references, I understand how heavy it can be to change things in that direction, so please consider this as
a question rather than an issue for the moment.

Laurent






-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.2.2 (MingW32)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQFEjqAAIF1tz5h+GDARAp8kAKCDpNp3fa2z2KsMEOGRj3VlUL7PlACcCGkP
gP8an5+MLUAz2a56oRBIyJw=
=EOmV
-----END PGP SIGNATURE-----

Received on Tuesday, 13 June 2006 11:22:47 UTC