[comp.dcom.lans] PC/IP and the Packet Driver.

dixon@gumby.paradyne.com (0000-Tom Dixon(0000)) (10/13/89)

In article <736@ftp.COM> jbvb@ftp.COM (James Van Bokkelen) writes:
> As of this moment, the following TCP/IPs for DOS support the Packet Driver
>and interface-sharing with BYU Netware (in order of introduction):
>
>FTP's PC/TCP (commercial)
>CMU version of PC-IP (public-domain; Karl Auerbach's TRW driver was the first)
>Phil Karn's KA9Q (no-commercial-use copyright)
>Clarkson version of NCSA (public-domain)
>WIN/PC from Wollongong (I'm not sure if this is commercially available yet).
>PC/NFS from Sun (not supported by Sun, but available via Clarkson).

A couple of questions:  
We have never tried this but have always wondered it. Can you use
the packet driver with multiple TCP/IP applications?  Say for example
NCSA Telnet and PC-NFS.  Or would the packet driver get really confused?


>If there weren't at least some people out there who think they're getting what
>they pay for, I'd be in another line of work.-- 

Good point.  Quality is worth something.  But sometimes, its sweat and not
cash.


>James B. VanBokkelen		26 Princess St., Wakefield, MA  01880
>FTP Software Inc.		voice: (617) 246-0900  fax: (617) 246-0901

Tom Dixon
AT&T Paradyne, Largo, Flj
uunet!pdn!dixon

romkey@asylum.SF.CA.US (John Romkey) (10/13/89)

In article <6646@pdn.paradyne.com> dixon@gumby.paradyne.com (0000-Tom Dixon) writes:
>We have never tried this but have always wondered it. Can you use
>the packet driver with multiple TCP/IP applications?  Say for example
>NCSA Telnet and PC-NFS.  Or would the packet driver get really confused?

No. The packet driver was designed to allow multiple protocol stacks
to share a network interface, and to hide the hardware details of the
interface from the protocol stacks. It was NOT designed to allow
multiple instances of one protocol stack run on top of it. Aside from
demultiplexing problems, there are all sorts of IP worms you're opening
yourself up for if you try to run multiple IP stacks on one interface.
-- 
			- john romkey
USENET/UUCP: romkey@asylum.sf.ca.us	Internet: romkey@ftp.com
"Live the life you love, Use a god you trust,
 and don't take it all too seriously." - Love & Rockets

jbvb@ftp.COM (James Van Bokkelen) (10/13/89)

In article <6646@pdn.paradyne.com>, dixon@gumby.paradyne.com (0000-Tom Dixon(0000)) writes:
> A couple of questions:  
> We have never tried this but have always wondered it. Can you use
> the packet driver with multiple TCP/IP applications?  Say for example
> NCSA Telnet and PC-NFS.  Or would the packet driver get really confused?

In the basic packet driver, the packet demultiplexing is done at the link
layer; by Ethertype in Class 1, or 802.2 header in Class 11.  Thus, there
can be only one stack getting IP packets (Ethertype 0x800) or XNS packets
(Ethertype 0x600) at any given time.  The second IP stack should get an error
on the access_type() call.  Some people at Clarkson hacked a special driver
to do demuxing at higher levels, but they may not be using it anymore.

My own opinion is that protocol stacks are too bulky to have two sets of code
doing the same thing (TCP or IP).  Our PC/TCP presents a standard programming
interface above the transport layer, as does PC/NFS, and I would suggest
porting the "Telnet and above" part of NCSA to use the other transport,
rather than introducing some IP-TPC demuxer that gets the packets where they
want to go based on the TCP port.
-- 
James B. VanBokkelen		26 Princess St., Wakefield, MA  01880
FTP Software Inc.		voice: (617) 246-0900  fax: (617) 246-0901

geoff@hinode.East.Sun.COM (Geoff Arnold @ Sun BOS - R.H. coast near the top) (10/13/89)

In article <6646@pdn.paradyne.com> dixon@gumby.paradyne.com (0000-Tom Dixon) writes:
>A couple of questions:  
>We have never tried this but have always wondered it. Can you use
>the packet driver with multiple TCP/IP applications?  Say for example
>NCSA Telnet and PC-NFS.  Or would the packet driver get really confused?

The packet driver interface that I shipped off to Clarkson was very basic:
it grabs all incoming packets rather than filtering on packet
types. Modifying it to do static filtering (i.e. all ARP, RARP, IP)
is pretty trivial.  (This is what's needed for coexistence with
NetWare or other non-IP stacks.) Modifying it for dynamic filtering based
on individual UDP or TCP ports, etc. is _much_ harder, and in
general not particularly useful: how, for example, do you persuade
NCSA Telnet to use a TCP port which doesn't clash with one
used by a PC-NFS Toolkit app?

Geoff Arnold,                              Internet: geoff@East.Sun.COM
PCDS Group, Sun Microsystems Inc.
---------------------------------------------------------------------------
"Who's next?" "Me, doctor?" "No, ME doctor, YOU patient." (Graham Chapman, RIP)

nelson@sun.soe.clarkson.edu (Russ Nelson) (10/14/89)

In article <738@ftp.COM> jbvb@ftp.COM (James Van Bokkelen) writes:

   In article <6646@pdn.paradyne.com>, dixon@gumby.paradyne.com (0000-Tom Dixon(0000)) writes:
   > A couple of questions:  
   > We have never tried this but have always wondered it. Can you use
   > the packet driver with multiple TCP/IP applications?  Say for example
   > NCSA Telnet and PC-NFS.  Or would the packet driver get really confused?

   In the basic packet driver, the packet demultiplexing is done at
   the link layer; by Ethertype in Class 1, or 802.2 header in Class
   11.  Thus, there can be only one stack getting IP packets
   (Ethertype 0x800) or XNS packets (Ethertype 0x600) at any given
   time.  The second IP stack should get an error on the access_type()
   call.  Some people at Clarkson hacked a special driver to do
   demuxing at higher levels, but they may not be using it anymore.
Right, we're not using it anymore.  We did it so that we could run RVD
and NCSA Telnet at the same time.  It worked, but it was kind of gross.

   My own opinion is that protocol stacks are too bulky to have two
   sets of code doing the same thing (TCP or IP).
I agree.
--
--russ (nelson@clutx [.bitnet | .clarkson.edu])
Live up to the light thou hast, and more will be granted thee.
A recession now appears more than 2 years away -- John D. Mathon, 4 Oct 1989.