On Thu, 16 Dec 1999, Jason Thorpe wrote:

> I'd like to include those unconditionally in all i386 kernels, but it
> would be nice to know if it breaks any currently-working systems, first.

one option would be to include them, and wait for people to complain. I
know this isn't a good blanket position, but we even bragged about this
particular code on the web page so it might be a good candidate for this.

for what it's worth, as a NetBSD user i would be happy if -current were
less stable than it currently is, if doing so would allow me to discuss
things and file pr's with a broader and more useful scope.  

especially now that -release gets lots of attention between releases and
people can get the release branch themselves fairly easily, and formal
releases happen with an oftenness that corresponds more closely with the
pace of hardware development, i think it's appropriate to be a bit
aggressive.  I think practically the only (user-based) consideration in
-current's stability should be maintaining an adequate audience of happy
bug-testers.

I, for one, have CVS locally and read on the web page that I should tag
my repository when I get a -current that works well.  I've never actually
done this, because I rarely get a -current that _doesn't_ work.

The typical nonworking cycle is not so bad. It's s'posed to be part of the
fun!

 1. Build new kernel.
 2. Boot new kernel--fails because New Feature X breaks Old Machine Y
 3. send-pr or post on the list
 4. Use old kernel.  Be happy.
 5. wait until next weekend.  Go to step 1.

The only time it's a pain is with ``latent'' brokenness.  I'm lazy and
poorly-organized enough that I can't easily downgrade if I find a problem
a week after building world.  But, that's a situation I can fix on my end.

-- 
Miles Nordin / v:1-888-857-2723 fax:+1 530 579-8680
555 Bryant Street PMB 182 / Palo Alto, CA 94301-1700 / US