[4738] in Kerberos
Re: NFS security & interoperability
daemon@ATHENA.MIT.EDU (Thor Lancelot Simon)
Sat Mar 4 17:05:05 1995
To: kerberos@MIT.EDU
Date: 4 Mar 1995 21:49:16 GMT
From: tls@cloud9.net (Thor Lancelot Simon)
In article <199503021649.LAA06999@gza-client2.aktis.com>,
Dan Geer <geer@cam.ov.com> wrote:
>
> That's nice. When there is a free source-code implementation available,
> I might think about using DFS in a production environment.
>
>You volunteering?
Not hardly. I don't really know much about DFS, I just object to running
stuff I can't get the source to in case I need it.
NQNFS seems to be okay, and I'm anxiously awaiting a stable release of
the free NFS V3.
Is DFS licensed like AFS -- X dollars per client, X dollars per server?
I've always tried to stay away from AFS because I very strongly believe
that networks with One Giant Server and lots of puny clients evince poor
network design. Caching client code and the like are nice, but the Right
Solution is to design one's network so that most file access will always go
to a local disk. To build that kind of network with AFS, I'd have to
pay for an AFS server license for every machine, which is prohibitive.
We looked at AFS while I was at Panix, because we had substantial
stability problems with the (alleged) NFS file locking code in SunOS and
with NFS write performance, but we had to reject it out of hand because
there was no reasonable pricing available for what we considered a
well-balanced network design.
It's kind of like what my boss told me when I came back from a
conference with a pile of Auspex brochures: "If we ever need to buy an
Auspex, you're fired." If DFS isn't licensed like that, I'll look at it
harder.
ObKerb: Does the NFS V3 code interoperate with the Kerberos support in
the stock 4.4BSD NFS? If not, this might pose a substantial problem for
me. Ick.
--
Thor Lancelot Simon tls@cloud9.net
Somewhere they're meeting on a pinhead, calling you an angel.
