[comp.protocols.nfs] Packet drivers vs. Non Packet drivers

GEustace@massey.ac.nz (Glen Eustace) (11/14/89)

I have been waiting for someone to start working with Packet drivers and
PC-NFS. This discussion would indicate that this is now happening.

We have about 200 PCs currently using PC-NFS 3.0.1 just about all of them
are using WD8003E clones from one source or another. My problem at this
time is that many people prefer to use the NCSA telnet rather than the
Sun supplied one. However, after using telnet one's PC generally needs to
be rebooted as the PC-NFS ethernet driver and NCSA telnet do not seem to
share the hardware very well.

The latest version of NCSA telnet we have will support a Packet driver
but as evryone knows Sun's PC-NFS officially does not.

Has anyone had any experience with Sun PC-NFS, the packet driver from
Clarkson ( courtesy of Geoff Arnold ) and the WD8003 card. Will it
overcome the problem with NCSA telnet, is it robust enough to warrant
installing it on 200+ machines?

Geoff? Is Sun going to include an 'official' packet driver with PC-NFS?

I would appreciate peoples comments.

-- -----------------------------------------------------------------------
  Glen Eustace, Software Manager, Computer Centre, Massey University,
   Palmerston North, New Zealand. Phone: +64 63 69099 x7440 GMT+12
             E-Mail via Internet: G.Eustace@massey.ac.nz
<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>

geoff@hinode.East.Sun.COM (Geoff Arnold @ Sun BOS - R.H. coast near the top) (11/16/89)

In article <388@massey.ac.nz> GEustace@massey.ac.nz (Glen Eustace) writes:
>I have been waiting for someone to start working with Packet drivers and
>PC-NFS. This discussion would indicate that this is now happening.
>
>We have about 200 PCs currently using PC-NFS 3.0.1 just about all of them
>are using WD8003E clones from one source or another. My problem at this
>time is that many people prefer to use the NCSA telnet rather than the
>Sun supplied one. However, after using telnet one's PC generally needs to
>be rebooted as the PC-NFS ethernet driver and NCSA telnet do not seem to
>share the hardware very well.
>
>The latest version of NCSA telnet we have will support a Packet driver
>but as evryone knows Sun's PC-NFS officially does not.
>
>Has anyone had any experience with Sun PC-NFS, the packet driver from
>Clarkson ( courtesy of Geoff Arnold ) and the WD8003 card. Will it
>overcome the problem with NCSA telnet, is it robust enough to warrant
>installing it on 200+ machines?

The packet driver is not a satisfactory way to achieve the
coexistence of multiple TCP/IP stacks (or multiple XNS, or...)
Suppose an ARP request comes in. Which stack should handle it? How
do the stacks avoid allocating duplicate port numbers? It is feasible if
the implementation can explicitly enable and disable packet
filters to select exactly the packets required (protocol, source
address/port, destination address/port). However the filter
mechanism in the Packet Driver is not sufficient to do this in the 
face of, e.g., IP options.

I wrote the PC-NFS Packet Driver driver (um.. well, you know
what I mean....) in order to handle Ethernet boards for which
Packet Drivers existed but for which we didn't have PC-NFS LLDK
drivers. The only coexistence which I envisaged was perhaps with
NetWare or some other non-IP stack. The code as it stands would
actually need a little work to do even this, but the sources are
available.

>
>Geoff? Is Sun going to include an 'official' packet driver with PC-NFS?
>

Glen, I honestly can't say at this point. Our current emphasis is on
investigating things like Microsoft/3Com's NDIS and Novell's ODLI.
MAC NDIS has an advantage over the Packet Driver in that each protocol
is offered a peek at the packet to determine algorithmically if
it wants it, and if not whether others should have a crack at it.
This is very similar to the way our drivers work internally (the
routine "demux" in the driver kit inspects the entrails and
pronounces "yea" or "nay"), which makes coexistence even at the
protocol stack level much easier.

To date, the only thing we have done with the packet driver is to
say to people "it's there, use it if you wish, but it's not supported".
Our legal and support people are concerned about the GNU-style
licensing and what that might involve if we did any more than this.
We have pointed a couple of customers with specific requirements
at the packet drivers: for example, a customer was doing some real-time
data gathering and wanted to be able to temporarily turn off the
Ethernet board interupts. He did this by hacking the source to the
corresponding packet driver: much more convenient for us (and him) than
us giving him the source to our native driver.

We recognize the importance of improving our telnet offering in 
PC-NFS. We are exploring a number of alternatives, and I am confident
that what we eventually come up with will meet the needs
of your users and many others. That's really all I can say
today. (I know, it's frustrating when vendors waffle like this,
but there really are good reasons for it sometimes.)

Geoff Arnold,                    | "This is not the 18th century. A bunch of
PCDS Group, Sun Microsystems Inc.| hicks with squirrel rifles can't provide a
Internet: geoff@East.Sun.COM     | good model for political action today."
Disclaimer: What he said.        | (Steve Cumming, stevec@theory.utoronto.ca)