RE: appinfo considered harmful. Re: What is the advantage of RELAXincomparison to Schemas?

["David E. Cleary" <davec@progress.com>]

>     Thanks for this clarification. I do think that an invitation
> to put all
> the features you would like to see but aren't  XML Schema via the appinfo
> mechanism is an invitation for problems. The introduction of  new
> "features"
> into XML Schema, in the hopes of such features making the next version, is
> asking for the browser wars revisited, and we are still smarting from that
> experiment.

I don't think these too things are similar. Nothing you put in appInfo can
stop you from processing the schema or an instance. It has more to do with
getting a job done. For instance, say you model up an appInfo vocabulary to
support co-ocurance constraints. This basically creates the notion of
"application valid" as opposed to "schema vaild". Any conforming processor
will be able to determine schema validity whether they know anything about
the appInfo. As for application validity, the application can either learn
to use the published appInfo vocabulary, or perform whatever validity
contraints the application requires and that is not expressible in XML
Schema. The use of appInfo allows you to keep this information with a schema
in a form that is machine processable instead of having it separatly in
prose for the application developer to implement.

But that is just one of the many uses for appInfo. It can be used for many
other things than just extra validation checking.

>     One of the reasons to have common specifications is to enable (not
> provide) interoperability. One of the major factors against
> interoperability
> is the use of custom slots/tags/whatever such as appinfo.

Again, nothing in appInfo stops you from schema validating an instance or
using its PSVI.

>     I'm not against modularization, its just that appinfo doesn't
> provide a
> standard mechanism to create and document such modules.

That is where the Schema Adjunct Framework comes in. Although it mostly
refers to a separate XML document, it will be mapped to appInfo structures
too. See http://www.extensibility.com/saf.

David Cleary
Progress Software