I'm confronted with the task of writing some code that will parse some XML and create a bunch of Python objects. e.g.:There are 35 classes, all with different sets of attributes, so I'd like to parse all these files without writing reams of parsing code for each class. I'm thinking of using dulcinea.specified to define all the expected attributes -- each class would subclass Specified -- and then use that information to convert strings to integers (or whatever) as needed. The only missing piece would be a convert_value(text, spec) which took the string 'text' and converted it to whatever Python type defined by 'spec'. Considering that it probably won't have to handle sequence types, convert_value() shouldn't be too difficult to write. Does this approach seem reasonable? Is there anything in dulcinea.specified that would make it a bad idea? Performance might be one reason, but 'specified' seems to stay nicely out of the way until you actually want to check an object against the specification. Hmm... while Googling this, I just turned up http://gnosis.cx/publish/programming/xml_matters_20.txt, which describes an approach that resembles dulcinea.specified quite a bit. Maybe this idea isn't as insane as I thought. --amk WackyName 45 ...