Need async-style exception handling.

Really nice library, very comprehensive and it ate up some very complex schemas that I work with from an industry data consortium.

One thing that would be really nice is if the marshall/unmarshall code used the normal convention of calling the callback with function(err, document). In the current situation, it is impossible to catch any errors in the serialization or de-serialization process. Document validation errors are far too common to just have the program crash every time it encounters one. If I'm reading the code correctly, this could be a simple matter of putting the try/catch block in each of the marshal/unmarshal messages and calling the callback with either the parsed document or the errors.

What would be really sweet, but a bit more work, would be to continue parsing in the face of all but catastrophic errors, and then call the callback with both the object and an error list.

thanks again,
~Bill