RE: [xml-dev] Bolt-in Type Systems (RE: [xml-dev] Relax NG and Web Services (formerly Joining the church))

["Dare Obasanjo" <dareo@microsoft.com>]

This is getting into implementation details of RELAX NG implementations of which I have little experience but it seems that this doesn't look like much of a problem. There are two options 
 
1.) A RELAX NG validator either supports certain types or does not. In this case if the datatype libraries are supported by the implementation then it's all good if not then an error is raised since the datatype library is unsupported. 
 
2.) A RELAX NG can dynamically learn how to validate a type library by dereferencing the namespace URI. I've heard there are Java interfaces type library authors can implement so arbitrary RELAX NG validators can support them. In this case there should be no problem as long as the dataype library is discoverable and implements the correct interfaces. 
 
NOTE: This is all conjecture and assumption on my part based on a passing knowledge of RELAX NG. 

________________________________

From: Bullard, Claude L (Len) [mailto:clbullar@ingr.com]
Sent: Fri 7/11/2003 12:30 PM
To: Dare Obasanjo; xml-dev@lists.xml.org
Subject: RE: [xml-dev] Bolt-in Type Systems (RE: [xml-dev] Relax NG and Web Services (formerly Joining the church))



What happens in an instance if it contains productions in different
namespaces where each has a RELAX NG schema but each schema references
a different data type library?  I assume the validator uses the different
definitions according to the namespace and does not merge these.  On the
other hand, the object model for the implemented system does.  I know this
isn't exactly clear, but I seem to guessing that any system, be it browser
or operating system, which manages multiple object models must support
the maximum set for all types referenced, thus the need for registries.
Extensibility proceeds exactly as one would guess: if the model wants
different types, it uses the URI to download and install them once
granted permission.  Following that example, namespaces would equal
semantics.

It may be time to read the NRL specification James announced.

XML is a software who-dunnit.  We wait for the next Clark work
as if it were from Rowling.  XML has become Harry Potter.

len


From: Dare Obasanjo [mailto:dareo@microsoft.com]

I think the problem depends on whether the schema will be used only for
validation or whether it will be used for type augmentation as well. There
is a big difference between check to see if this number is prime
(validation) and the datatype of this number is prime number (type
augmentation). One is feasible to implement in modern systems while the
other is not.

As for datatype libraries being in conflict, I guess I need to see an
example scenario before I can confirm or deny that it is possible. Some
would probably consider the ability to create union types and ambiguous
grammars as potentially creating datatype conflict. However I am not sure
this is what you mean.