Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!zaphod.mps.ohio-state.edu!swrinde!cs.utexas.edu!sun-barr!newstop!eastapps!shark-bb!geoff
From: geoff@shark-bb.East.Sun.COM (Geoff Arnold @ Sun BOS - R.H. coast near the top)
Newsgroups: comp.protocols.nfs
Subject: Re: PCNFS(tm) and Clarkson Drivers
Message-ID: <6549@eastapps.East.Sun.COM>
Date: 5 Jun 91 13:15:51 GMT
References: <3267@dsacg2.dsac.dla.mil>
Sender: news@East.Sun.COM
Reply-To: geoff@east.sun.com (Geoff Arnold @ Sun BOS - R.H. coast near the top)
Organization: Sun Microsystems PC-NFS Engineering
Lines: 40

A few comments on Jim's posting.

(1) The product is actually PC-NFS(R), not PCNFS(tm). >sigh<

(2) For PC-NFS 3.5 you should add the "/s" switch to the PCNFS.SYS line
throughout. Sorry about that.

(3) Jim asks (rhetorically) why anyone would want to bother with
running PC-NFS over packet drivers.  There are a couple of reasons he
didn't mention.

First, and obviously, the packet driver may be the only usable driver
for your board.  For example, we don't ship a driver for the NE1000,
and the NDIS driver for this adapter is (as far as I can determine)
badly broken. The packet driver works very nicely.

The second reason is if you plan to switch between PC-NFS and some
other software which uses the packet driver (e.g. SOSS): it certainly
helps your configuration if you can use the same driver in each case.
(If you have to switch i/o addresses, for example, you only have one
configuration line syntax to worry about. In the case of the 3C503, the
PC-NFS driver requires that shared memory be disabled, while the packet
driver uses shared memory, so you have an extra incentive to use a
common driver!)

Finally, if you have the time and the inclination it may be worthwhile
trying all of the driver options (native PC-NFS, packet driver, NDIS)
to see which is smallest/fastest. You should not assume that our native
driver is automatically going to be the best (particularly for those
boards for which we use vector drivers from 3Com). Different drivers
perform differently on different hardware: for example, if you have a
system with a fast CPU but relatively slow DMA, using a driver that
employs shared memory or programmed i/o will probably be a win. Since
drivers rarely advertise the way they do i/o, you'll have to
experiment.

-- Geoff Arnold, PC-NFS architect, Sun Microsystems. (geoff@East.Sun.COM)   --
------------------------------------------------------------------------------
--     Sun Microsystems PC Distributed Systems ...                          --
--            ... soon to be a part of SunTech (stay tuned for details)     --