[xsd-users] Questions on few features.

[Balasubramanyam, Shivakumar]

Thanks for detailed respons. I have removed text I agreed with and have no more questions. My other comments are inline.

________________________________

> 4.    This is more to solve above point in a more generic way. Has
> there been any thoughts on decoupling the C++ objects that represent the
> XSD datatypes separate from feature supports like serialization or XML
> parsing etc.

There is a way not to generate pretty much every aspect of the code that
may not be needed. For example you can use the --suppress-parsing option
to get rid of the XML parsing code. In fact if you pass --suppress-parsing
and don't pass --generate-serialization, you will end up with just the
object model that can neither read nor write XML.

However, I am not sure how useful this feature is to you. If I understood
you correctly, you still want to create XML documents but without using
Xerces-C++. It seems to me that what you really want is a way to plug a
different XML serializer instead of Xerces-C++ (BTW, which one would
you prefer?). This is not possible with the C++/Tree mapping but we
are working on another mapping similar to C++/Tree that will allow
changing the underlying XML parser/serializer.



[Shiva] I am using the xerces parser and have no issues at present. It seems like you guys are already headed

towards other serializers. I was thinking plugging in a binary serializer. Like Boost, the serializers could be XML, binary or simple text logging.

But XSD has the important function to generate the mapping and structure. Why should XSD depend on xerces for logging in XML?

If I have a binary protocol that I would like to use schemas to define, then I could use XSD to generate code for binary serializing and deserializing

but log the messages in XML for easy debugging.

Thanks,

Shiba

 

hth,
-boris