On Fri, Jun 20, 2003 at 04:23:12PM +1000, Charlie Allom wrote:
> On Fri, Jun 20, 2003 at 06:08:59AM +0000, Sean Davis wrote:
> > current state of affairs (especially on SMP machines, from what I read on
> > the lists) counts as stable by any definition of the word.
> > 
> 
> This is the same argument as the UFS2 checkin.

Indeed.

> 
> -current is for bleeding edge. The SMP code hasn't been fiddled with
> much because the nature of the NetBSD project doesn't assign anyone to
> tasks they must do..

True. But note, however, that before the UFS2 checkin and the nathanw_sa
merge, current hadn't had any severe bugs that I noticed for years. The only one
I can think of since 1.5 branched was a thunderlan driver bug which caused
occasional kernel panics. This was fixed within a couple months by
Christos.

I don't consider that bug to be as severe as the pthreads or UFS2 bugs for
three reasons:
1) it only happened every few days or so. 

2) it didn't interfere with crash dumping, so I could at least get debugging
information, which IIRC helped Christos find the root cause of the problem
and fix it.

3) it did not cause the loss of thousands of files.

Let us not forget, also, that either -current or a
post-sommerfeld_i386mp-but-pre-nathanw_sa snapshot is the ONLY choice for
those who want NetBSD SMP on i386 boxes. I would not be surprised if that
fact alone is driving users away from NetBSD and towards FreeBSD, which
already has stable pthreads and SMP.

> I am hurting because I put -current on my production boxes. It is not
> stable. I've not got anyone to blame but myself though :)

I realize that I could be running netbsd-1-6, but -current is preferable to
me because it, well, gets the neat new stuff first. I could just wish that the
neat new stuff went through more testing before getting thrown into the tree.

-Sean

-- 
/~\ The ASCII
\ / Ribbon Campaign                   Sean Davis
 X  Against HTML                       aka dive
/ \ Email!