RE: [xml-dev] Backward and forward compatible schemas ... Relax NG -->

[noah_mendelsohn@us.ibm.com]

Rick Jelliffe wrote:

> What people complain about with XML Schemas 1.0 is having to adjust the
> base schema when they want to just be able to adjust the derived
> schemas. (That you have to add wildcards or <other> elements willy nilly
> is as old as the hills, and no advance on parameter entities or
> <redefine>.) 
> 
> However, the whole basis of XML Schemas is namespaces and modularity,
> and that means adopting standard or industry vocabularies which almost
> universally have not been written to be open. People are really loath to
> rewrite standard or industry vocabularies.

Yes.  In particular, Schema 1.1 not only loosens the UPA rule that made it 
difficult to use wildcards with optional content, it provides at the type 
level and the schema document level the ability to say "validate as if 
there were a wildcard either (a) at the end of my declared content or (b) 
everywhere that I might be about to reject content while validating.  You 
get to say what those wildcards are, so it's relatively easy to say things 
like:  "I allow open content everywhere in this type, but that content 
must not be from namespaces N1 and N2; or, you can say, that content must 
be from my namespace;  or you can say "it can be anything other than the 
following elements.", etc.  Not perfect, but seemingly useful.

> that being said, the new features in XSD 1.1 for openness look a 
> step in the right direction for capabilities

Glad you think so.  I know you've had problems with schema in this regard.

> except for adding to the monolith and thereby being a step in the 
> wrong direction complexity-wise.

Yes, an important concern.  The language is and always has been too big 
IMO.  As to whether it's conceptually well layered I think parts of it are 
pretty good on that score, and parts are a mess.  As you know Rick, the 
one aspect of it all that I think was probably the right tradeoff, and for 
which I take some personal responsibility, is that conformance is pretty 
much all or nothing, the goal of course being that the same schema works 
the same way everywhere (admittedly the unnecessary complexity of the 
language has somewhat undercut that goal, but I think that having 
all-or-nothing conformance on the stated semantics has been an aid to 
getting it widely adopted.)  Anyway, I'm glad you feel that the "openness" 
features may be effective.  Thank you.

Noah


--------------------------------------
Noah Mendelsohn 
IBM Corporation
One Rogers Street
Cambridge, MA 02142
1-617-693-4036
--------------------------------------