Subject: dnscache (was Re: Presence of one disk trashes another)
To: Frederick Bruckman <>
From: Sean Davis <>
List: tech-userlevel
Date: 11/20/2003 21:23:53
On Thu, Nov 20, 2003 at 07:39:53AM -0600, Frederick Bruckman wrote:
> On Thu, 20 Nov 2003, D'Arcy J.M. Cain wrote:


> You'll see that most of the requests are stalled until after 1.6.2 is
> cut, as it's evidently very close, so there's no point in rushing in a
> pull-up request. Once it's made, you can always pull-up the identified
> changes to your own sources in the same way that releng does, if you
> don't want to track "netbsd-1-6".

Since I haven't seen the code get pulled up, is 1.6.2 going to have the same
broken resolver code netbsd-1-6 currently has? (ie, works fine with bind,
doesn't work at all with dnscache?) It's been fixed in -current for some
time now, but nobody seems interested in fixing it in -1-6, so thats a big
disadvantage to using the 1.6 branch, considering I (and others) use
dnscache and like the ability to resolve things. :)

I understand that 1.6.2 is going to happen soon, and that only critical
bugfixes are being put in at the moment, as far as I know, but this is
critical - not fixing that bug is akin to deliberatly being incompatible
with djbdns. I know a lot of people hate djb for his licensing or for other
reasons, but refusing to fix broken code (that worked before it was broken,
of course) is nothing short of introducing an intentional incompatibility,
especially when the problem is known and fixed in -current, and involved
simple changes to one file.

I won't suggest a pullup, the last time I did that I was flamed to death by
multiple NetBSD developers, but I will suggest someone think long and hard
about whether 1.6.2 should be released with broken libc resolver code.

Consider: someone using dnscache & netbsd 1.6{,.1} upgrades to 1.6.2. suddenly
they can't resolve anything using dnscache anymore. That doesn't seem like
good PR for NetBSD to me :)


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