Subject: Re: /etc/rc vs /etc/init.d, take N
To: None <current-users@NetBSD.ORG, mouse@Collatz.McRCIM.McGill.EDU>
From: Peter Seebach <email@example.com>
Date: 04/10/1996 23:08:54
>So? As with the POSIX braindamage, just because someone else does it
>is not, in itself, a reason you/we should do it.
I disagree. Standardization is, in and of itself, a good thing, and
being like other systems is, in and of itself, a reason to do something.
It is not always a good reason, or a sufficient reason, but there *is*
a benefit there. (It may easily be outweighed by the losses of a bad
idea. Consider deciding to go single-tasking "for compatability with
DOS programming style.")
The degree to which standardization is worth pursuing depends on the scope
of the standard, and the applicability. POSIX compliance, and ANSI/ISO C
compliance, are both relevant and respectable standards, and compliance
with them is desireable.
By contrast, compatability with 4.3BSD is (in some cases) an utter waste
of effort. (Let's have a show of hands from everyone who would *really*
prefer to have (char *) sprintf.)
None of this is particular to this given case; it applies to nearly any
issue involving extensions, features, or unimplemented features.
(As an example, I'm designing a set of enhancements for printf() for
consideration by C9X, and one of the things I'm doing is examining
existing implementations to see what *they* do. I won't do the same
things in some cases, but it's a good source for ideas. I will go
out of my way to avoid having a function with the same name and
radically different semantics.)