Backward-compatibility of ntpq may be an issue

Eric S. Raymond esr at thyrsus.com
Fri Dec 23 13:12:10 UTC 2016


Hal, I was reviewing some recent changes, and it occurred to me that
your commit changing the default of showall in ntpq may have not
been a good idea.

I don't think the showall functional change in either ntpmon or ntpq
was a bad thing in itself.  But backwards-incompatible changes that
might affect scripting worry me some. I think of the NTP userbase as
very conservative and likely to get upset if "ntpq peers" in a
pipeline does not have stable behavior.

Security concerns give us a lot of permission to make incompatible
changes.  Outside of those, I think we have only a limited budget of
tolerance to spend.  I have tried hard not to not fritter away that
budget on unimportant changes, so I would have it to spend on larger
ones. Sometimes this concern has blocked me from cleaning up and
simplifying UIs, with ntpq itself being a particular case in point.

I don't think the showall change in ntpmon is a problem; it's a new
tool, there's no inertia.  ntpq is a different matter.

I think you are as careful about this sort of thing as I am when
it is brought to your attention, and I'm serious about delegating
control to people who take responsibility, so I'm not going to
revert that change myself.

Instead I'm going to ask you to seriously consider doing so,
for the reason I have given.

A general discussion of when backward compatibility matters
would not be amiss here.
-- 
		<a href="http://www.catb.org/~esr/">Eric S. Raymond</a>

"Those who make peaceful revolution impossible 
will make violent revolution inevitable."
	-- John F. Kennedy


More information about the devel mailing list