Subject: Re: running -current in production (was: libpthread)
To: Richard Rauch <rkr@olib.org>
From: Greg A. Woods <woods@weird.com>
List: current-users
Date: 06/21/2003 14:37:53
[ On Saturday, June 21, 2003 at 07:07:55 (-0500), Richard Rauch wrote: ]
> Subject: Re: libpthread
>
> -current has always been known to me as the "version" that is not even
> guaranteed to *compile* on a given day.  It compiled?  It compiled!
> Great.  So it crashes, what did you expect?

The problem is that there are often so many highly desirable features in
-current, and such a long time between formal releases, that running
at least some snapshot of -current in production is quite common.  I've
got customers running -current, but of course it's not just a shapshot
downloaded from releng.netbsd.org, but rather a very carefully tweaked,
tested, and fixed up version of the code that I snarfed on a day when I
had heard rumours that it not only compiled but would also run.  I.e. I
created my own private release from -current for customers to use in
production.  The unified buffer cache was the primary feature that made
such effort worthwhile.  As of netbsd-1-6 I've been happy enough to stay
on the release branch, though SMP (_without_ libpthread :-) would be nice.

-- 
								Greg A. Woods

+1 416 218-0098;            <g.a.woods@ieee.org>;           <woods@robohack.ca>
Planix, Inc. <woods@planix.com>; VE3TCP; Secrets of the Weird <woods@weird.com>