Re: RA with MV attributes
Date: Sat, 20 Jan 2007 22:26:07 -0700
Message-ID: <MPG.201ca75e6be62d789897d2_at_news.altopia.net>
Marshall <marshall.spight_at_gmail.com> wrote:
> In fact, if I understand
> Rice's theorem, *all* nontrivial properties of code
> are undecidable.
I'll be the picky person who points out that Rice's Theorem says all non-trivial properties of the (languages/problems/functions) that are (accepted/solved/computed) by code are undecidable. Clearly, the question "does this program use 'foo' as an identifier?" is both nontrivial and decidable.
But yes, I agree with what you are ultimately saying. I'm being picky for the following reason. I would say that the distinction of properties of code versus its language/problem/function is relevant precisely because it explains that some kinds of analysis are possible in general even for a Turing-complete language, but they will just never correspond exactly to a property of the language. What can happen is that, for example, the failure of analysis to prove correctness doesn't NECESSARILY mean that a program is incorrect, but is a pretty damn good indicator in real life such that it would be a remarkable coincidence if the program actually were still correct.
-- Chris SmithReceived on Sun Jan 21 2007 - 06:26:07 CET