Copy Link
Add to Bookmark
Report
TCP-IP Digest Vol. 1 No. 02
TCP/IP Digest Wednesday, 14 Oct 1981 Volume 1 : Issue 2
Today's Topics:
Administrative Trivia
NCP-to-TCP Transition Deadline!!
Performance of 3Com "UNET" UNIX TCP/IP Software
Behind the sceens at 3Com & Stanford PUP
Final word on BBN PDP-11 TCP/IP Implementation
RFCs and IENs -- How to get to NIC
Gateway questions
----------------------------------------------------------------------
From: Michael Muuss <mike@bmd70>
Subject: TCP-IP Digest, Plans, etc.
At this point, everybody who has requested to be added to the
TCP-IP discussion list has been added. In addition, Jon Postel had a
list for distributing notices about Internetworking changes. At his
suggestion, I have incorporated that list into the Digest list. If you
are reading this, and don't want to see any more of these Digests,
please sent a note to TCP-IP-REQUEST@BRL.
The scope of the Digest will probably exceed the rather specific
"TCP-IP Digest" title, but that is OK by me. I see this as a forum for
discussing implementation and design problems relating to large scale
networks, and internetworking. Anything too practical for HUMAN-NETS
is fair game here (at least for now). But I would hope that discussion
will focus on IP and TCP, because this is where much of the real action
seems to be.
At the present, I expect to be able to issue one digest per week,
with each issue being (hopefully) 15000 to 30000 characters long,
depending on submissions. No more mamouth 55K digests (OOPS!). If the
interest picks up, I can see being able to do as many as 3 per week.
More than that, and I will have to get help, but it will be no problem.
Already, some people have mentioned that mailing stuff to
"TCP-IP@BRL" might be hard to remember or type. I will be glad to set
up as many aliases for the list as seem useful. Everybody: please
suggest some mnemonic, short aliases!
I am keeping an archive here on BRL (0/29), and hope to announce
the method by which they can be retrieved, sometime soon. (However,
we can't permit anonymous login, so it will probably be strange). If
other sites wish to keep an online copy that can be FTP'ed, please let
me know, and I will summarize for the list.
Cheers!
-Mike
------------------------------
From: POSTEL at USC-ISIF
Subject: NCP-to-TCP Transition
It is really very important for everyone to notice the deadline for
completing the cutover to IP/TCP and the elimination of NCP from use
in the ARPANET.
The deadline is: 1 January 1983.
That is 14 and a half months from now. Really not much more than a year.
--jon.
------------------------------
From: croft at SRI-UNIX
Subject: UNET at SRI
The SRI Telecommunications Sciences Center has been running the 3COM
UNET software (release 1.5) on two 11/44's for the past few months
and we have been quite pleased. This is contrary to the false rumor
circulated in the last TCP/IP digest (V1N1). Both machines run a standard
V7; installation was very smooth and straightforward, requiring no
hacking of the kernel. We chose the 3COM software over BBN's "free"
V6 TCP for just this reason. One of 11/44's is on the ARPANET as
SRI-PRMH (packet radio measurement host). A locally written device
driver for UNET provides TCP/NCP multiplexing of the single 1822
interface. The other 11/44 is currently being used in a packet radio
experimental network at Fort Sill, Oklahoma. The device driver in the
latter case speaks CAP5 protocol over an 1822 interface to the radio.
Throughput of the ARPANET version seems to be quite good (compared to
the V6 BBN): 20K bits/second on an FTP to ourselves with the packets
looping all the way out the 1822 and back in again. Of course UNET
is doing twice as much work here as normal and one would expect
perhaps 40K bps if two UNETs were FTPing thru the same IMP.
Although UNET is certainly the fastest PDP11 TCP around, it still has
the thruput constraints inherent in any TCP. (See the excellent
article by Bunch and Day of DTI: "Control Structure Overhead in
TCP"). 3COM claims 100K bps max thruput for their TCP on DEC hardware.
I think a previous digest entry mentioned 50 to 60K average thruput on
DA11 hardware (DA's are about 8Mbps devices). My beef is that with
simpler protocols one should be able to do much better. As an example
I mention the UNIX network at Purdue University EE department.
It uses 1M bps DMC and DMR link hardware and achieves 400K end to
end process usable bandwidth. Perhaps TCP's on a fast local network
should support some simpler subset of the protocol as an option to
allow improved thruput.
--Bill Croft (croft@sri-unix)
------------------------------
From: Bill Nowicki <CSD.NOWICKI at SU-SCORE>
Subject: Re: Re: TCP/IP for UNIX
I know Bob Metcalfe, Ron Crane, Bruce Borden, and Greg Shaw either directly
or indirectly; these are the main movers behind 3COM. Bob invented Ethernet
at Xerox, with Ron designing some of their controllers. Bruce and Greg are
Unix wizards, so between them there is a lot of expertise. 3COM has been
targetting their marketing to small systems (11s and even Onyxes), and since
we (Stanford) have mostly Vaxes and 20s, we are pressured into going with
BBN's implementation because it was Arpa-sponsored.
By the way, we are currently using the PUP Internetwork architecture at
Stanford, which Metcalfe had a hand in designing. It was quite a step
for him to adopt the DOD Internetwork architecture for the sake of
standardization. A welcome change from the "not invented here"
syndrome.
-- Bill
------------------------------
From: Michael A. Wingfield <wingfiel at BBN-RSM>
Subject: TCP/IP
Mike,
I have a report which documents the [BBN] unix tcp/ip which I can
send you. If you want it, send me your mailing address.
We have not done any work on the software for about 2 years,
so it is probably the same as your 115 tape. We have no
plans to do anything else to the 11/70 version of tcp/ip.
Thanks,
Mike
------------------------------
From: Zellich at OFFICE-3 (Rich Zellich)
Subject: Please add me...
...to the TCP-IP distribution. Are you aware of all the stuff on TCP
and IP in the IEN documents maintained by the NIC (similar to the RFC's,
but relating to the Internetwork Experiment funded primarily by DARPA and
taking place primarily on the ARPANET)? I suggest that at least some mention
of the IEN's be made in an early issue/message for anyone who is interested
but may not yet know about them. I don't know of any general distribution
list for notice of new IEN's (I'm on distribution for the IN Experiment group,
though not participating), so many people may not yet be aware of them.
Cheers,
Rich <ZELLICH at OFFICE-3>
------------------------------
[ Here is a repeat of a letter from the V1#1 which describes some RFCs
availible at SRI-NIC. -MJM ]
From: Postel at USC-ISIF
RFCs 791, 792, and 793 define IP and TCP. The ARPANET already
supports these protocols. Many TOPS20s, Unix-es (including VAX)
and MIT-Multics, and UCLAs IBM system, already have these protocols
in use. Work is in progress to replace all TIPs with TACs that
use IP and TCP. There are about 10 internet gateways in service
already. ARPA has set a goal for the complete switchover to IP
and TCP by January 1983.
------------------------------
From: James.Gosling at CMU-10A (C410JG40)
The mail standards are documented in RFCs (Requests For Comment) from
the network information center at ISI. To get a copy of an RFC just ftp
to NIC, log in as user "anonymous", password "guest" and retrieve the file
<NETINFO>RFCnnn.TXT . nnn is the RFC number, some useful ones are:
RFC733 Mail message format standard
RFC754/RFC799 Discussions of internet addressing
RFC759 Internet mail transfer protocol
RFC780 Mail transfer protocol
James.
------------------------------
From: lou at UCLA-Security (Lou Nelson)
Subject: New List
I would very much like to be on the TCP/IP mailing list.
I think a good first entry would be the report you compiled
that everyone wanted to see. If it is not available, perhaps
you could do a short summary if it's not too much work.
I'm getting a VAX up at Aerospace Corp. that will run UNIX
in December and currently the only good prospect for getting
it on the net is the BBN TCP/IP that Berkeley is pounding on
as a final debugging measure. I don't understand the way
(which sites) I get to an NCP host from the TCP VAX yet.
Pointers to any RFC or other document that explains that will
be appreciated.
Regards,
Lou Nelson
[ Please reply to TCP-IP@BRL, so all can benefit from any answers -MJM ]
------------------------------
End of TCP-IP Digest
********************
