The HPI Seaside tutorial shows the data models as subclasses of Object, and I have seen that elsewhere. However, some examples talk about finding the closest class to use as your starting point. I am curious what experienced Squeakers do.

Subclassing Object presumably gives maximum flexibility, but perhaps means that one ends up building in functionality that already exists further down the tree. On the other hand, taking a more specific class, might be a bit like premature optimisation?

Cheers
AB