> Are you saying that the allowed arguments of the isValid() method on the
> Datatype interface depends on the return value of the isContextDependent()
> method on a separate interface? That doesn't seem right to me. The Datatype
> interface should be semantically self-contained. I think the
> isContextDependent() method must be on the Datatype interface. This isn't
> so bad, as it's potentially useful for other applications independently of
> the Datatype compatibility spec. The getIdType() can be on a separate
> CompatibilityDatatype interface.
Then I think it's OK to have all methods (including getIdType()) on the
Datatype interface.
Implementing the getIdType() method is easy (just return ID_TYPE_NULL)
and so it's very easy for datatype libraries to support it.
If they are separate, processors need type checking and casting to call
the getIdType method. I guess less optional features are better.
We've added the createStreamingValidator method, which is arguably an
optional feature, into the Datatype interface. So I think it's OK to
treat the getIdType method in a similar way.
regards,
--
Kohsuke KAWAGUCHI +1 650 786 0721
Sun Microsystems
