On Fri, 29 Oct 2004 15:25:26 EDT, David Brodbeck said: > This suggests that it's reasonable for a program to segfault because the > user made a mistake, instead of having some non-fatal form of error > handling. I don't think that should be acceptable at all, though I agree > it's very common. If I had a dollar for every time I've lost work because a > segfault or GPF happened before I saved my document... The problem is that if you say "it isn't acceptable at all", then you *VERY* quickly end up with almost *NO* software that's "acceptable". All software is buggy. *ALL* of it. Learn to accept it. Making a "this is not acceptable" declaration about something that in general is not totally preventable is just doomed to failure. Also, remember that programmer time is *FINITE*. Would you have been willing to not have the last 3 "this is indispensable" features in your favorite software in order that the programmers track down every single possible failure mode? Would you give up 2? 1? Most software projects could probably fairly easily eliminate (rough guess) some 75-90% of the really bonehead errors via better programming methodology and automated software testers/verifiers - but after that, it's going to be *really* hard to get much further improvement. There are no silver bullets....
Attachment:
pgpg0mvpMEMmx.pgp
Description: PGP signature