TCP-IP@Brl.ARPA (04/20/83)
TCP/IP Digest Wednesday, 20 Apr 1983 Volume 2 : Issue 4
Today's Topics:
TCP/IP -vs- Xerox NS
Merging Protocol Sets?
MIT-CSR TCP/IP on non-I/D UNIX
3Com UNET on non-I/D UNIX systems
BBN TCP/IP on non-I/D UNIX systems
DEC Equipment & TCP/IP
----------------------------------------------------------------------
TCP/IP Digest --- The InterNet Digest
LIMITED DISTRIBUTION
For Research Use Only --- Not for Public Distribution
----------------------------------------------------------------------
Date: Tue, 12 Apr 83 10:34 PST
From: Taft.PA@parc-maxc.arpa
Subject: TCP/IP vs Xerox NS
To: TCP-IP@brl.arpa
Let me suggest that there is not much point in engaging in a "TCP vs.
NS" debate, since there's far more to discuss than can reasonably be
presented in this forum. People who want more information about NS
would be better served by reading the open literature (references below)
and forming their own judgements.
The Xerox NS family of protocols are a re-engineered version of the Pup
protocols. NS and Pup are true internet protocols, architecturally very
similar to TCP/IP. This is not accidental, as there was considerable
cross-fertilization between the Xerox and ARPA internet projects.
Of course, there are also many differences. Some are a reflection of
differing technical judgements made by the designers, while others are
of no significance. But the fundamental similarities are far more
important than the detailed differences.
So why is Xerox pushing NS rather than TCP/IP? Well, why do similar but
incompatible products appear in any marketplace? Xerox has a big
investment in NS, and had to make product plans based on it long before
TCP/IP had solidified to its present form. NS is not really "new".
Though the protocol specifications were published only a little over a
year ago, NS has been around for several years, and its predecessor,
Pup, has been in active use since 1975.
Ed Taft
References (there are many others; these are the ones that come
immediately to mind):
D. R. Boggs, J. F. Shoch, E. A. Taft, R. M. Metcalfe, "Pup: an
internetwork architecture", IEEE Transactions on Communication, vol.
COM-28, no. 4, April 1980. (This was a special issue on network
architectures and protocols; it contains many interesting papers.)
Y. K. Dalal, "Use of multiple networks in the Xerox Network System",
IEEE Computer magazine, 15(10), October 1982.
J. E. White, Y. K. Dalal, "Higher-level protocols enhance Ethernet",
Electronic Design, 30(8), April 1982.
------------------------------
Date: 12 Apr 1983 0823-PST
Subject: TCP/IP vs. XNS
From: Dan <LYNCH@Usc-Isib.ARPA>
To: tcp-ip@Brl.ARPA
ISI has decided to purchase a bunch of Xerox 8010 workstations
and will be implementing TCP/IP on them in the next 6 months.
We will be in a position to comment on the relative merits
of each protocol suite after we have our TCP/IP implementation
running. The first task we face is to figure out where to make
the cut. We will have enough access to the low level drivers
to be able to make the cut anywhere we choose, but we are not
sure that is the best way to "merge" the functionality of
these two "similar" protocols. The whole insanity of "layered"
protocols becomes quite apparent when one considers what we are
about to delve into -- one from culumn A and one from column B
and then two from column A, etc???
Dan Lynch
------------------------------
Date: 12 Apr 1983 1026-EST (Tuesday)
From: lwa@Mit-Csr.ARPA
Subject: Re: Paul Milazzo's TCP request
To: TCP-IP@Brl.ARPA
Our TCP/IP implementation is running on one LSI-11/23 that I know
of. It's pretty tight, though, and certain pieces (like the
Internet reassembly code) had to be omitted. It's going to
take some amount of work on the part of whoever brings it up
to get it to fit.
Regarding the earlier user-only implementations (eg. BBN's): we
actually had the BBN implementation running on an 11/40 at one
point, but there were only three disk buffers left, so it wasn't
very useful...
regards,
Larry
------------------------------
Date: 12-Apr-83 10:32:43-EST
From: gc@Bnl.ARPA
Subject: TCP/IP for nonsplit I/D systems
To: milazzo.rice@Rand-Relay.ARPA
Cc: tcp-ip@Brl.ARPA
We are currently running the 3Com Unet implementation of TCP/IP on a v7
system on an 11/44, however there are installation directions in our manual
for installing on non-split I/D systems and on BSD 2.8 systems also.
>From our experience, it might be a tight fit but probably worth a try.
------------------------------
Date: 12 Apr 1983 0943-PST
From: DEDWARDS@Usc-Isi.ARPA
Subject: re' TCP/IP for UNIX LSI-11/23
To: milazzo.rice@Rand-Relay.ARPA
cc: tcp-ip@Brl.ARPA
As it turns out I am running TCP/IP on my 11/34 and 11/23 - this
is the external to the kernel version done at BBN several years ago.
The underlying UNIX is a BBN version 6, but it should be movable
to V7. It does require that you have portions of the old NCP
buffering mechanism installed though, so you do need to have
room to add stuff to your kernel (also needs Rand ports, await/
capac, pty, etc). With only a couple of users and full memory,
the 11/34 does pretty well on the ARPAnet - the 11/23 is somewhat
slower.
Howard Weiss
DoD Computer Security Center
------------------------------
From: smb%mhb5b@brl-bmd
Date: 13 Apr 83 11:26:51 EST (Wed)
Subject: TCP/IP for UNIX LSI-11/23
To: unc!milazzo.rice@Rand-Relay.ARPA, unc!tcp-ip@brl-bmd.arpa
Full-Name: Steven M. Bellovin
I'm pretty sure that 3Com's UNET will run on 11/23s, under either v7 or 2.8BSD.
A source license is about $5000, I believe.
------------------------------
Date: 14 Apr 83 20:51:19 PST (Thu)
From: unisoft!billn@Ucb-Vax.ARPA
Subject: tcp on 11/23
To: milazzo.rice@Rand-Relay.ARPA
Cc: tcp-ip@Brl.ARPA
UNET from 3com runs on an 11/23. At least they say so; I've never tried it.
I suspect it does, since they developed much of UNET on 11/23's. There are
alot of options you can turn off to get sizes down in user programs. Kernel
size is a bit of a problem, but solvable, using the kernel overlay scheme
("kov's") from Ken Harrenstein (klh@sri-nic) The new version (UNET 2.0) is
reputed to be up-to-date arpa compatable, including smtp. They include,
naturally, a driver for their q-bus ethernet controller.
/bill
------------------------------
Date: 23 Nov 1982 1653-PST
Subject: DEC equipment and IP/TCP
From: Joel Goldberger <JGoldberger@USC-ISIB>
To: TCP-IP@BRL
I can tell you what the situation is regarding IP/TCP implementations on
most DEC equipment. There are basically four operating systems that
people run on DEC 10/20's and two operating systems that are run on
Vaxes.
On the 10/20's people are running:
TOPS-10
TENEX
TOPS-20
and ITS (The MIT Incompatible Timesharing System)
BBN has had an implementation of IP/TCP for TENEX and TOPS-20 for some time
and that is what we are running. Very few other sites were willing to run
this software though. DEC proposed a much cleaner user-interface to TCP for
TOPS-20 that most of the TOPS-20 sites decided to wait for. This code was
originally scheduled to be made available to ISI at the beginning of July,
since then the date has been slipping. Monday the person working on it at
DEC (Kevin Paetzold) announced over the TOPS-20 distribution list that he
would be making the code available on 1 December. Obviously once the code
is delivered there will be some lag before the support software gets written
and debugged, and I seriously doubt that all of that can be accomplished in
the one month before the switchover.
I don't believe that anything besides the present BBN implementation of
IP/TCP will ever be available for TENEX systems, and most of the TENEXes on
the Network already have this code up and running. There is some work
needed to make the support programs run under TENEX, but I believe Henry
Miller at NIC is doing this.
Ken Harrenstein has been (re-)hired by MIT to implement IP/TCP on the ITS
machines (MIT-AI/DM/ML/MC). I believe he is already back at MIT doing this.
I know of no implementation for TOPS-10 (or WAITS), either available or in
development.
On VAXes people mostly run either VMS or UNIX (primarily Berkeley UNIX).
For VMS there are two implementations available:
Digital Technology Incorporated offers a comercial product called ACCESS-T
in a Binary only distribution that includes all the usual servers and user
programs (FTP/TELNET/SMTP), as well as a library of user callable routines
for establishing and controlling IP and TCP connections. We ordered this
product for our VMS system and were given a very early Beta-Test version
that was essentially unusable (It crashed VMS every time a TELNET connection
was closed). They did some debugging on our system and fixed most of the
problems, but we still chose not to run it for the time being. This
decision was mostly due to the way they had implemented MAIL; sending and
receiving were two disjoint programs and there was no way to REPLY or
FORWARD messages conveniently. They have just announced a new version of
their software that will run under version 3.0 of VMS; The previous release
would not. We will be evaluating this latest version when we bring up
version 3.0 on our VAX.
David Kashtan at SRI has been working on porting the Berkeley 4.1aBsd (UNIX)
implementation of IP/TCP to run under the EUNICE UNIX compatibility package
that he authored and SRI distributes. (We are currently running his ported
version of the Berkeley NCP code on our VAX.) I received a message from him
on Friday (19 November) that he had succeeded in bringing up IP/TCP, and was
now running his VAX (SRI-IU) with both NCP and TCP. We will be getting a
copy of his code in the next week or so and hope to use that on our VMS
system. I assume he will make it available to anyone who wants it.
Under Berkeley UNIX there are also two choices:
BBN (Rob Gurwitz) has an implementation for Berkeley 4.1Bsd which many sites
are running (including us). It comes complete with user and server
processes, source code, and again a library of user-callable routines for
establishing connections. We have found it to be very stable and Rob is
very good about bug-fixes and general support. I don't know if he plans on
offering an implementation for Berkeley's future releases.
Berkeley (Bill Joy) essentially re-wrote the BBN implementation for
Berkeley 4.1aBsd, and the soon to be released 4.2 Bsd. We have been running
this version on VAX 11/750's connected to a 10MBit EtherNet and have found
it also very stable. 4.1a & 4.2 include a set of network utilities that
allow access to remote file-systems and remote command execution, features
not directly available with the BBN implementation.
I hope this has been of some help to you.
- Joel Goldberger -
------------------------------
END OF TCP-IP DIGEST
********************