[2269] in Kerberos
Re: Kerberized NFS in 4.4BSD
daemon@ATHENA.MIT.EDU (Steve Dyer)
Thu Oct 8 05:13:08 1992
Date: 8 Oct 92 05:14:38 GMT
From: dyer@spdcc.com (Steve Dyer)
To: kerberos@shelby.Stanford.EDU
In article <9210071559.AA01256@snowhite.cis.uoguelph.ca> root@SNOWHITE.CIS.UOGUELPH.CA (gopher I) writes:
>- The specification of the MIT mechanism is not public either. My
> understanding is that MIT logs all requests for the info. for Sun and that
> this suggests that Sun considers some of the provided code/info proprietary.
This comes as a surprise to me; Ted, comments? Aren't diffs available?
The standard NFS3.2 OEM source is licensed from Sun. But I'd imagine all the
MIT changes would slide into a public-domain UNIX NFS implementation smoothly
(or at least be obvious to someone who knew what they were doing.) I can't
imagine that any of the MIT code is proprietary--we're talking extra case
statements in the rpc.mountd dispatch switch statement. The credentials map
stuff in both rpc.mountd and the kernel is completely self-contained and
completely new code.
You need a kernel hook into rfs_dispatch() or the equivalent to implement the
credentials mapping of the RPC packets before any particular NFS operation
is performed, and a way to add the nfsmapctl() system call to your kernel.
This isn't rocket scientist material.
>It might also be observed
> that the MIT mechanism is not ideal, but simply the best they could do
> without making NFS protocol modifications.
Yep, purely mountd protocol additions, and a conservative approach to ensure
interoperability. In hindsight, it would have been prudent to have made the
effort to reserve these "official" rpc.mountd calls with Sun.
--
Steve Dyer
dyer@ursa-major.spdcc.com aka {ima,harvard,rayssd,linus,m2c}!spdcc!dyer
