Subject: Re: Thanks for help with DS5000/125
To: Aaron J. Grier <agrier@poofy.goof.com>
From: Jonathan Stone <jonathan@DSG.Stanford.EDU>
List: port-pmax
Date: 10/27/1999 12:05:33
In message <19991027115148.F4507@goldberry.poofy.goof.com>",
Aaron J. Grier" writes:

>I have no problem mounting NFS exports from my 2.2.12 linux box to my
>NetBSD/pmax machines:

I'm sure you don't.  OTOH I've had terrible trouble doing NFS mounts
from Linux-2.2.12+knfs to NetBSD 1.4.1 machines, both i386 and pmax.

Removing the RH 6.0/6.1 knfs and going back to Mark Shands' unfsd (or
whatever the Linux guys renamed it) works, but at the cost of any
claims to good performance.


>The big difference with my setup is that I'm not running the knfsd
>shipped by redhat, (it's known to be broken,) but one compiled from
>sources provided by H.J.Liu, including rpc.knfsd and its ilk, as well as 
>patches for recent kernels which for some reason haven't gotten back
>into the offical Linus' releases.

Could you email a pointer to these to current-users? There's enough
Linux crap around that it's causing problems for a number of us.  And
I see a realaudio proxy in the 3.3.3 code, which Bernd is going to
import into -current over the next week.

Or did you want a back-port of 1.3.3 to the 1.4 -release branch?
(NB: you could always grap ip_fil3.3.3 and run the BSD/kupgrade script,
 but there are some NetBSD-specific fixes in the NetBSD tree, and I
 havent  checked  that they are all folded into 3.3.3)


><RANT>Of course, this, being Linux, within the last year or so all my
>old references for a stable NFS under linux have all but dissapeared or
>moved without much hints.  Linux' distribution and packaging scheme is a
>horrible fragmented mess.  If I didn't already have so much time
>invested in Linux, and NetBSD could do some of the things Linux can in
>the NAT area [1], I'd be temped to dump it entirely.  Thank god for
>NetBSD.</RANT>

  ``linux software engineering is an oxymoron'' -- anon.

Still, if you could let us know exactly what Linux features NetBSD
cannot do, that'd be interesting.  ipnat seems to work very well for
many users; what's it missing that you need?


>[2] which of course has been rewritten for the third time within the
>    last couple months, meaning I need to now learn yet another set of
>    tools.  Grr...


See above about ``linux software engineering''.