Carlo Pescio - blog

Although it is not widely recognized, object oriented programming brought several concepts from artificial intelligence, in particular from [definitional] semantic networks.
I'm old enough to have taken an Artificial Intelligence course :-) back at the university, where we studied semantic networks among other things (truth maintenance systems, non-monotonic reasoning, frames, logic programming, etc). Overall, quite interesting stuff, but it didn't deliver. One of the reasons, in my opinion, was the huge waste of energy spent in the pursue of completeness, not to say perfection. Semantic networks, for instance, had a concept of inheritance, and lot of time was spent to accommodate misclassification, like the birds that can't fly. Robert Pirsig (author of "Zen and the Art of Motorcycle Maintenance" and "Lila") calls those hard classification problems the platypuses of a theory, and AI spent too much time tinkering with platypuses.
When OOP came along with the notion of inheritance, the whole misclassification problem was largely ignored (leaving aside a few academic papers). The OOP community basically took the useful stuff, said "who cares" about the rest, went on and delivered. They moved the problem of misclassification from the concerns of the system to the concerns of the developer. We can violate the LSP at our own risk. We do our best to avoid that, and meanwhile we build useful systems. OOP does not pretend to be a complete, perfect system; that's why it can do useful work.
Today, Aspect Oriented Programming is an interesting paradigm that seems to be spending a little too much time with its own platypuses. It's just a matter of time before someone will come along, grab the low fruits, and deliver. .NET context attributes (plus transparent and real proxies) and java refection proxies, are not powerful enough, but they're a first step away from all the aspect weaving blurb and toward a simpler interception model, and in a few cases they can actually deliver. We just need something easier and more powerful :-), and that would be well within our reach, if only simplicity was a little more valued.
To my completeness-and-perfection-oriented readers: remember that Goedel, in 1931, proved that in any consistent formal system, which is adequate for arithmetic, there are propositions that cannot be proved or disproved within the axioms of the system (that's the famous Goedel Incompleteness Theorem, informally). So all our beautiful math, all our revered logic, is not complete either. That's why it can do useful work.