Re: Unnecessary change

[Richard Stallman]

    By unifying this, we can prevent these problems in the future - if there is
    a bug, we can fix it at one place instead of hunting why we now use three
    forms to include the same file:

    pop.c:#include <../src/config.h>
    profile.c:#include <config.h>
    sorted-doc.c:#include "config.h"

If we were using all three forms in different places, then it does
seem reasonable to converge on one.  But we should choose one that has
been in use for a long time.  The log shows that sorted-doc.c only
started to include config.h in 2000, which means users have only seen
it since 21.1 was released.

Meanwhile, the log says that profile.c was changed by you on Oct 8.
In Emacs 21.1, it uses #include <../src/config.h>.

I think we should converge on a method that has been used since Emacs
20.  That way we will know we can have confidence in it.  If multiple
methods were used in Emacs 20, any of them is ok with me.

I've learned from long experience that "don't change it if it works"
is a very wise approach for dealing with a wide variety of systems,
many of which I have never seen.  No matter what generalities you
think you know about these systems, there could be exceptions.
"Nobody has complained about this in ten years" is not perfectly
reliable, but it is the most reliable basis that we ever have
to be confident that something works on all systems.


We raised the issue a few weeks ago of which kinds of changes we
should discuss in advance, rather than just installing them.
Experience shows that changes that affect the mechanisms that deal
with variability between systems are the most likely to lead to
long-delayed problems.  From now on, could we please discuss in
advance any such change?