On Wed, 18 Oct 2000 13:41:34 -0700, Greg Stein <gstein@lyra.org> said:
>
> A comment from the peanut gallery... I *really* dislike environment
> variables which modify the operation of commands. They can lead to too many
> surprises.
>
> "why is it doing THAT?! ... search ... search ... oh, dear lord! damn
> fuggin environ variable? die die!"
>
> No way. Dunno what two-bit thought they would be Goodness, but let's
> *please* rethink the whole proposition of *why* environ variables can or
> should be used. Personally, I find zero reasons.
I agree with the sentiment, but env variables can be a helpful evil,
although one could argue not a necessary one.
For example -- say I'm trying to compile some crummy non-
autoconfiscated piece of software to go into flash ROM, and I really
need it to compile with `-m486 -Os'. Either I can go and modify
several dozen manual invocations of the C compiler (yuk) or I can set
CFLAGS, which has some hope of Doing What I Mean without too much work
on my part.
That is to say, the environment variable CFLAGS allows me to cope with
the fact that the author of the software has bogusly failed to give me
a convenient hook for modifying how the C compiler is invoked, as I
would have if the software were autoconfed.
Accordingly, I think that your point dovetails very nicely with
Jonathan Shapiro's distinction between knobs that are really
descriptive of the *project* (such as what generated files to ignore),
and knobs that are a matter of personal/local preference (such as the
degree of verbosity).
I offer that the latter class of knobs should be specifiable in the
environment.
Received on Sat Oct 21 14:36:11 2006