And in addition, for a long time the problems we were solving mapped well onto the kinds of decompositions we can do easily/naturally with functional/procedural (de-)composition.

In fact, if I had to choose one and only kind of connector, that's the one I would choose.

Dataflow expressed functionally or procedurally is pretty awful, but it has nothing on the horrors of fibonacci expressed using dataflow. Yes you can do it, but you really don't want to.

AFAICT, it is exactly because the connector we have is so good that we haven't really noticed how limiting it is to have just the one. A kind of gilded cage, or as I call it: The gentle tyranny of call/return[1]. And I really do mean the "gentle" part.

And for some weird reasons, whenever people proposed different connectors, they almost invariably proposed them as a single connector to replace all others. Which never turns out well.

[1] https://2020.programming-conference.org/details/salon-2020-p...