3 ms·
I'm sad that in this thread and https://news.ycombinator.com/item?id=11786193 https://news.ycombinator.com/item?id=11786193, nobody is actually talking about th
by Ericson2314 10y ago
I'm sad that in this thread and https://news.ycombinator.com/item?id=11786193 https://news.ycombinator.com/item?id=11786193, nobody is actually talking about the technical arguments at play. I've only read the EWD692 "opening salvo" but already there are interesting things to point out.
EWD1303:
> The profound significance of Dekker's solution of 1959, however, was that it showed the role that mathematical proof could play in the process of program design. Now, more than 40 years later, this role is still hardly recognized in the world of discrete designs. If mathematics is allowed to play a role at all, it is in the form of a posteriori verification, i.e. by showing by usually mechanized mathematics that the design meets its specifications; the strong guidance that correctness concerns can provide to the design process is rarely exploited. On the whole it is still "Code first, debug later" instead of "Think first, code later", which is a pity, but Computing Science cannot complain: we are free to speak, we don't have a right to be heard. And in later years Computing Science has spoken: in connection with the calculational derivation of programs —and then of mathematical proofs in general— the names of R.W. Floyd, C.A.R. Hoare, D. Gries, C.C. Morgan, A.J.M. van Gasteren and W.H.J. Feijen are the first to come to my mind.
Backus as quoted in EWD692 (Djikstra attacks this):
> One advantage of this algebra over other proof techniques is that the programmer can use his programming language as the language for deriving proofs, rather than having to state proofs in a separate logical system that merely [sic!] talks about his programs.
From a perspective of a PL person like myself, these ideals are very compatible. Without correctness by construction, there is not enough proof reuse so formality will forever be doomed for niche applications (aka computers embedded in dangerous things). Likewise after trying out intuitionistic type theory--based things (e.g. Agda) separating the program and proof langauges just seems clumsy. Overall separating programming and proof whether spatially (seperate languages) or temporally (correctness proved after the fact) is bad, and the reasons hardly depend on the dimension of separation
- Ericson2314 10y agoIIUC, Backus also talks about correctness proofs with "Von Neumann model" languages just being a lot harder to prove, a separate point from the language for the proof itself. Well, that's a huge selling point for PL research to this day, Dijkstra evidently came fully on board years later as evidenced in https://www.cs.utexas.edu/users/EWD/transcriptions/OtherDocs/Haskell.html https://www.cs.utexas.edu/users/EWD/transcriptions/OtherDocs...: > A fundamental reason for the preference is that functional programs are much more readily appreciated as mathematical objects than imperative ones, so that you can teach what rigorous reasoning about programs amounts to.
- pron 10y ago> separating the program and proof langauges just seems clumsy. Lamport's TLA (which serves as the basis for TLA+) is simpler than Agda because in Agda, even though you use the same syntax for both types (i.e. propositions) and programs, they are conceptually separate, while in TLA they are the same. Both the program (well, its specification) and its properties are just logical propositions. TLA+, which is untyped (and based on TLA and ZFC) has the added advantage of being far, far (far) easier to learn than any dependently-typed language. TLA was introduced in this paper[1], which begins thus: > Correctness of the algorithm means that the program satisfies a desired property. We propose a simpler approach in which both the algorithm and the property are specified by formulas in a single logic. Correctness of the algorithm means that the formula specifying the algorithm implies the formula specifying the property, where implies is ordinary logical implication. We are motivated not by an abstract ideal of elegance, but by the practical problem of reasoning about real algorithms. Rigorous reasoning is the only way to avoid subtle errors in concurrent algorithms, and we want to make reasoning as simple as possible by making the underlying formalism simple. However, there is a debate between the PL approach (championed by Backus, Milner) and the specification approach (championed by Dijkstra, Lamport) over utility. The question is whether proving at the code level is empirically viable (at a reasonable cost) for all but the simplest of programs. Indeed, so far the answer seems to be no. There has been only one program ever written and verified using dependent types (CompCert), it is certainly non-trivial but rather small, yet it required a world-class expert, took a lot of effort, and even then required corner-cutting (Leroy says he gave up on proving termination, simply adding a counter and throwing a runtime exception if it runs out). TLA+, OTOH, is used extensively and regularly by Amazon (and Oracle, Microsoft and others), by "plain" engineers, on real-world large systems (far larger than CompCert), and management loves it because it actually seems to save them time and money. This is only possible because TLA+ is not the programming language. Lamport doesn't reject the theoretical possibility of a PL that could provide large-scale verification of some sort (end-to-end verification will always be extremely expensive due to simple complexity arguments), only points out that no PL has so far come anywhere close to fulfilling this promise. Of course, TLA+ doesn't provide end-to-end verification, but that is too expensive -- and unnecessary -- for 99% of the industry anyway. I think that the ideas are compatible, but so far in theory only, while in practice they are not (yet?). [1]: http://research.microsoft.com/pubs/64074/lamport-actions.pdf http://research.microsoft.com/pubs/64074/lamport-actions.pdf