3 ms·
Many of the ideas you discuss here were formalized in the 70s by Amir Pnueli and others into the LTL and CTL temporal logics -- ideas that still form the core o
by pron 10y ago
Many of the ideas you discuss here were formalized in the 70s by Amir Pnueli and others into the LTL and CTL temporal logics -- ideas that still form the core of much of software verification today -- and those heavily influenced both Lamport's temporal logic TLA and specification language TLA+, and the synchronous language Esterel, by Gérard Berry[1] (Wikipedia says that the synchronous language Lustre was influenced by Lucid) -- "time doesn't exist between stable states". Both TLA+ and Esterel have achieved a certain degree of success in the industry (to this day), although mostly in safety-critical realtime software and hardware design; the ideas are just starting to flow to "everyday" software. While they have a very explicit, clear notion of time modality and are "pure", they are not functional.
I think that as many ideas are crystallized, it is no longer helpful to characterize every pure and/or temporal-logic approach as functional; indeed, the two notions may often be at odds. (Pure) functional programming, at least as (sort-of) defined today. FP is not directly about modeling time and certainly does not impose synchronicity (a clock), but is more about the idea of denotational semantics, where computations are imagined as functions, and function application imposes an equivalence relation (the function application is said to be equal to its result in this relation). This is very different from the philosophy of synchronous languages.
[1]: The modern language Céu is based on Esterel: http://ceu-lang.org/ http://ceu-lang.org/
- alankay1 10y agoThanks for the additional references (Yes, I should have mentioned Amir's work also). A good basic point is that the central ideas of "reasoning with time" go back into the early 60s and -- unfortunately -- did not get into the mainstream thought patterns of everyday computing. Terms often lose their meanings as they become religions and more rigid choices and styles. The two in question here are certainly "object-oriented" and "functional programming"!
- pron 10y ago> Terms often lose their meanings as they become religions and more rigid choices and styles. The two in question here are certainly "object-oriented" and "functional programming"! Absolutely, especially when they're vague. But I still think that the synchronous or temporal-logic based languages (which I love!) are very much distinct from FP (at least in the Haskell "tradition"), although they share with each other -- and not with the classical imperative approach -- the explicitness of state transitions: the synchronous/TL model with its clocks/steps, and the FP model with its (hideous) monads. They also "feel" very different (Esterel/Céu and even TLA+ feel more imperative than functional). It is not surprising that the killer-app for the FP approach is compilers (and other code analysis tools), while the killer-app for synchronous/TL approach is reactive systems, two application classes that are at the opposite, extreme ends of Pnueli's (I think) classification of programs (transformational, interactive and reactive).
- alankay1 10y agoI don't identify the general notion of "functional programming" with any particular language, including Haskell. I stay with the "idea of a function" as a reliable mapping. Similarly, I don't identify the general notion of "object-oriented" (as I characterized it long ago) with any particular language. Both of these terms have now been "colonized" and their meanings changed, etc.