[an-allen]

I mean, I’m actually somewhat curious how starting tertiary computer science curriculum with formal methods would impact the students. I mean I felt my intro undergrad 101 course was essentially functional programming and lambda calculus (with Scheme)- and I felt that helped establish a fundamentally different way of thinking about computer science than a more basic procedural type introduction would have. When they moved the intro course to be OOP instead I felt it was a travesty.

[ht_th]

I did follow a curriculum set by Dijkstra's students. In the first trimester, we learned to program in Pascal, so we knew what programming was. The next two programming courses, in the second and fourth trimester, "programming" meant proving programs correct. Using pen and paper. Often in one-on-one sessions with our teachers where we'd to demonstrate our proofs. Or the teacher would state a problem and then continue to derive a correct progam by construction, writing slowly on the blackboard. These programming courses were supplemented with logic, discrete mathematics, and other formal methods. And we had continuous mathematics courses as well, practical labs, electronics, databases, and whatever you'd expect in a computer science and engineering curriculum.

For most students this wasn't easy, particularly compared to the way most of them were comfortable programming on their own by trial-and-error hacking away at a problem. Proving programs correct by construction takes a different skill. At the same time, it wasn't particularly hard either once you got going.

I don't think this way of teaching and learning programming was very useful or practical. With Dijkstra's students leaving the university, or otherwise losing primacy at the computer science faculty, Dijsktra's ideas faded away from the curriculum. Since then—I returned twenty years later to teach at this university—, the curriculum is very like any other computer science / engineering curriculum. And students seemed to have as much trouble with it as before.

What I missed about the curriculum when it was gone, was the consistency it brought into the curriculum. The curriculum felt as one continuous track to some clear idea of what it meant to be a programmer in Dijkstra's style. If you liked that idea, the curriculum was a great guide. If you didn't, it felt as a waste of time.

[mrkeen]

> compared to the way most of them were comfortable programming on their own by trial-and-error hacking away at a problem.

This is only a viable strategy insofar as the tool which lets them hack away is itself correct. You need computer scientists producing these tools.