3 ms·
I have no problem with "proofs are programs", only with "programs are proofs", and I think that most of theoretical computer science refuses, almost on principl
by pron 10y ago
I have no problem with "proofs are programs", only with "programs are proofs", and I think that most of theoretical computer science refuses, almost on principle, to view computation merely as programs expressed in any particular formalism, just as physicists would say that a certain ODE is a description of a falling apple, not that a falling apple is an ODE, nor that an apple can only fall if we can describe it in some formalism.
See the discussion here about realizers vs proofs: https://news.ycombinator.com/item?id=13495547 https://news.ycombinator.com/item?id=13495547. Perhaps you can shed more light on the subject. Harper's post is interesting, but while it does offer less narrow a view than Wadler's using a different definition of proof (which I don't think deserves the name proof, but nevermind), it is still too narrow a definition of computation (and his introduction of types seems to muddy the water and complicate matters further, partly because the use of types necessitates yet another computation, because types are certificates). Computation is a wider notion than constructive mathematics (as every construction is a computation but not every computation is a construction, and two different computations can yield the same construct).
As to the choice of a foundation of mathematics, I think this is largely an aesthetic matter. It is certainly a separate discussion from computation in the sense that computation does not depend on the choice of mathematical foundation, nor that the choice of mathematical foundation must be built around the notion of computation. I like Turing's view that there are no privileged foundations, and a mathematical formalism should be chosen based on utility, on a case-by-case basis.