Subject: Re: CVS commit: src
To: Andrew Brown <atatat@atatdot.net>
From: John Hawkinson <jhawk@MIT.EDU>
List: source-changes
Date: 02/22/2003 00:17:06
Andrew Brown <atatat@atatdot.net> wrote on Sat, 22 Feb 2003
at 00:05:34 -0500 in <20030222000534.B2577@noc.untraceable.net>:

> To: John Hawkinson <jhawk@netbsd.org>
> Cc: source-changes@netbsd.org

Same deal here; please don't discuss things on source-changes
extensively. Moved to tech-userlevel.

> >Log Message:
> >Use $diff_options when running diff in /etc/security.
> >Default diff_options to -u, for unified-format context diffs,
> >because context is essential to a useful evaluation of differences.
> >This represents a behavior change.
> 
> diff_options should default to empty to preserve existing behavior.
> now that there's an option, individuals can set it to something else
> of their own volition.

Please see the discussion in the PR (17247).

But again, no. We are not 100% bound by existing behaviors.
It is acceptable to make behavioral changes that are a net win.

In this case, it is clearly beneficial to provide contextualized
diffs over traditional context-free diffs. It is highly unlikely
that anyone will be inconvenienced by the change, but that minority
may easily reset diff_options="".

The best time to make the change to the default is now, when the
flag is being added.

If the change is high-visibility, which I don't really think this is,
then a heads-up can go to tech-userlevel and/or current-users.
I don't think this change merits such notice or warning, but
in any event, here we have it.

There is not much to be gained by adding useful options if they are
never on by default. Few users (out of the many who would benefit)
take the time to go through the complete change list and find
the options that should have been on by default in a sane system.
Backwards compatibility at this level is an impediment to be avoided,
not a religion to be slavishly adhered to.

--jhawk