[qmtest] RFE concerning 'Extension' class

[Stefan Seefeld]

Mark Mitchell wrote:

> I suspect that we would be better off by adding new useful test classes,
> or by making documentation improvements, as several have suggested. 
> Yes, a more elegant system would be easier to document -- but would it
> be *that* much easier to document?  And what about having to keep the
> backwards-compatibility code around for the indefinite future?

Yes, I think that's the critical point: how much backward compatibility
do you need / want ? If all has to remain as it is now (in particular
the convention about a 'arguments' variable holding a list of Field
instances) a new implementation using a metaclass would indeed not
contribute much.
So I reply with a question: Do you think it is feasable to migrate
(at whatever pace is possible for you) from the current convention
to the one I suggest ?

Regards,
		Stefan

PS: I'll try anyways to reimplement the current Extension API using
     a metaclass, just to see how it would look...