[This article was originally posted on August 17, 1995, in the midst of a thread on whether it's a good idea to use void main() and the like.]

Newsgroups: comp.lang.c
From: scs@eskimo.com (Steve Summit)
Subject: Re: Why the warning???
Message-ID: <DDGMr2.My3@eskimo.com>
X-Scs-References: <199508171432.HAA14633@mail.eskimo.com>
References: <00001a80+00003293@msn.com> <dqua.808547526@dqua> <40s7l2$mcr@garuda.csulb.edu>
Date: Thu, 17 Aug 1995 14:54:37 GMT

Earlier in the thread, someone expressed a popular sentiment: "I use what works." Raise your hand if you've ever said this. (My hand's up.) Now raise your hand if you've ever spent hours going through some code weeding out gratuitous nonportabilities so that you could get it to compile under a compiler or operating system other than the original author used. (My hand's still up.)

There's always a tradeoff to be made between portability and convenience. Almost no code is truly portable, even code that tries very hard to be, so it's reasonable to accept certain nonportabilities in one's code, in the interests of expediency. But unilaterally saying "I use what works" is an extremely risky attitude. It's one thing to make an informed decision: to weigh the costs of being strictly portable against the risks of using something nonportable. It's quite another thing to declare that you can't even be bothered to think about the tradeoffs, and to throw yourself and your code (and perhaps your company) at the mercy of future compilers you'll want to use.

Ask yourself: how long do you expect your code to last? If it's a quick hack that you don't expect to use past next week, then most of this is moot. But if it's production program, a product, that you or someone depends on or makes money on, you probably hope (you ought to hope) that it will be successful and long-lived. And it's likely, even though you can't imagine it today, that some day you'll find yourself wanting to compile it under some completely different compiler, or for some completely different machine or operating system. And, when that day comes, the more your attitude during development has been "I use what works," the more miserable the porting task will be.

When, for the platform and compiler you're using today, there's no alternative to a given bit of nonportability (for instance, some operating-system-dependent interaction which C doesn't define), there's no point spending lots of time desperately trying to make it portable. You can spend the time porting that bit of the code when and if the time comes. (You might spend a bit of time today making sure that the nonportable bit is cleanly isolated.)

When, on the other hand, there's a perfectly valid, equally easy, portable alternative, such as using int main() in preference to void main(), or /* */ comments in preference to // comments, it really seems foolish not to avail yourself of as many Standards as you can.

You can't imagine what kind of compiler you might suddenly find yourself wanting to use a year from now. Neither can I. But one thing we do know is that if it is a C compiler, it is extremely likely to comply as closely as it can to the ANSI/ISO C Standard. So the more closely we can make the code we write today conform to that Standard, the fewer problems we are likely to have down the road.

void main() and // comments aren't perfect examples, because // comments (for no good reason that I know of) are likely to be in the next major revision of the C Standard, and void main() works under most compilers and is likely to remain so (even though it doesn't have to) if for no other reason than that so terribly much code uses it. One could, I suppose, make an informed decision to keep using those two, if one really thought they were preferable to the strictly-portable alternatives. But in the more general case -- for the other errors and nonportabilities your code is all too likely to contain if all you care about is that it works today -- someone, someday, is likely to curse you if you've made their life miserable, or to sing your praises if you've expended a modicum of thought and effort towards giving your code a fighting chance of being portable down the road.

Steve Summit