Talk:Proof assistant

Older comments
See also (merge?) Computer-assisted proof Outs 10:05, 12 September 2007 (UTC)

Human help

 * "a human can guide the search for proofs"
 * Can someone give an example where there is a practical need for a human to guide the search for a proof? What kind of proof steps are better done by humans? --Abdull (talk) 08:54, 11 September 2010 (UTC)

Agda

 * I'm not quite sure where to put it in this list, but Agda probably belong here, too. KLuwak (talk) 14:20, 2 April 2011 (UTC)

Automath
The first proof assistant - it's still around and should be included. Here's the raw data. Others make look into it, in greater depth, and decide whether to include it. Name: Automath. Version: 4.2. Developer(s): Freek Wiedijk (implementation), N.G. deBruijn (specification and definition of the language and its variants). Implementation Language: C. Higher-order logic: Yes. Dependent types: Automath originated the concept. Small kernel: Yes. Proof automation: No. Proof by reflection: ?. Code generation: AUT-68 and AUT-QE are already executable. AUT-ΔΠ can be generated, but is not executable. Link: (http://www.cs.ru.nl/~freek/aut/index.html). We may branch off of this and put up our own version ("Version 4.3") in the near future. — Preceding unsigned comment added by 2603:6000:AA4D:C5B8:222:69FF:FE4C:408B (talk) 00:38, 24 February 2021 (UTC)

Small Kernel
I have experience of using Agda and Coq, and as far as I know their source code neither of them has a small kernel. --Konstantin.Solomatov (talk) 05:15, 20 July 2014 (UTC)

Comparison table: contents and meaning?
It would be great to add the logic language and rules used by the system (even if some PA like Isabel can deal with more than one set of rules). — Preceding unsigned comment added by Horaceb (talk • contribs) 10:38, 5 July 2016 (UTC)

I edited the table where I thought some of the feature evaluations were unclear. What are "partial" dependent types, for instance? There either are or aren't some types dependent on specific values. So far as I know, that's the definition of dependent types. If someone wants to put "partial" back, could they please explain what it means? Similarly for proof automation, ACL2 and PVS---which do not have small trusted computing bases---were listed as having "partial" automation because the automation is unverified; i.e., it is in the TCB. That seems already addressed in the Small Kernel section, to me.

This is a minor improvement, but I think overall it's not clear why the specific set of features in the table were chosen, or what exactly constitutes "partial" or N/A. Nothing in the table has a citation, for instance, to show that some reputable source both considers these criteria significant and agrees with the results presented here.

MisterCarl (talk) 22:23, 20 September 2012 (UTC)

Idris
Maybe someone wants to add Idris. I don't know all the fields to fill. 91.66.12.246 (talk) 19:19, 29 December 2016 (UTC)

More columns
It would be nice to see new columns for: --Jordan Mitchell Barrett (talk) 21:12, 12 December 2020 (UTC)
 * Whether the software is free / open-source / collaborative / etc or not
 * The programming paradigm (declarative / imperative / etc)

Discussion of proof checking/verification has been dropped!
"Automated proof checking" redirects here, but there is no substantive discussion of automated proof checking, merely mention of this as an aspect of the functionality of specific proof assistants. "Proof verification", mentioned in other articles, also redirects here. This is a serious problem. The content here is gone.

History
It would be nice to have a section giving a brief history of the term "proof assistants" and how it developed.

(I'm looking at the page now because I'd never heard the expression before. I'm retired from CS and Math and wondered when the name was coined.)  — Preceding unsigned comment added by 2607:FEA8:1BA0:A300:A5DB:51FB:64C9:32A9 (talk) 14:13, 27 November 2021 (UTC)