Excerpts from Nina Jeliazkova's message of Tue Sep 29 15:42:19 +0200 2009:
> You mean even if we have multiple implementation of dataset services,
> the conversion functionality should be restricted to e.g. dataset service?
> 
> The initial dataset proposal actually does assume conversion facilities,
> simply by being able to return different formats by specifying MIME
> types. For example, a dataset with format SDF is POST-ed, but then can
> be retrieved with format CML (or ARFF, YAML, etc.)
> 
> Do you think there is a need for specific "conversion service" rather
> than relying on Content-type ?

I would not add it to the next API version. A separate conversion
utility (maybe an extension of the reporting service) could make sense
in the future, if we discover that a lot of services are implementing
the ѕame methods for reading/writing external formats.

> 
> > Just to confirm my own understanding: There is no "standardised" XML way
> > to (de)serialise common datastractures - I need to know the schema to
> > reconstruct a datastructure from XML.
> >
> >   
> Well, no, the purpose of XML is to serialize domain specific objects,
> not necessarily data structures. The reason is to provide implementation
> neutral description of domain objects, which is quite flexible IMHO (the
> same set of molecules and properties might not be implemented as hash
> table in a third party service, yet serialized to the same data format).

Thanks! Thats the reason, why I have to define my own builders/parsers
with most XML libraries (I am addicted to the convenient to_yaml and
from_yaml methods).

Best regards,
Christoph