fred@mot.UUCP (Fred Christiansen) (12/06/84)
[]
Thanks to everyone who graciously mailed me their experiences in
dealing with controllers and transceivers of differing Ethernet types.
Since several requested that I post a summary to the net, here goes:
Motorola Microsystems contracted with a 3rd party to develop and manufacture
VERSAbus- and VMEbus-based Ethernet controllers and to port another vendor's
software plus XNS into a controller + 68000/68010-based std Unix Sys V
environment. At one point, the vendor supplied us with new controllers
(using LANCE instead of emulator board) AND a new software package.
When reliability was not up to snuff we first suspected software but then
discovered the problem was with mismatched controller/transceiver combos.
One can acquire controllers and transceivers of either type "Ethernet 1" or
"Ethernet 2/IEEE 802.3". As long as controller and transceiver match in
type, you do not have this problem. Moreover, these differing type combos
can both coexist and communicate over the Ethernet. For TCL transceivers,
model numbers with an "I" or "IS" suffix are of type 802.3 and those with
model numbers with an "E" or "EB" suffix are of type Ethernet 1.
Here now are the responses:
------------------------------------------------------------------------------
From: oakhill!ut-sally!seismo!elsie!cecil!keith
Date: Thu, 1 Nov 84 15:39:27 EST
To: oakhill!mot!fred
Subject: Re: Ethernet Transceivers
There are actually THREE types of Ethernet transceivers, version 1.0
(e.g. TCL's 2010EC), version 2.0 (e.g. DEC's H4000), and IEEE 802.3
(e.g. TCL's 2010I?). As I understand the situation, these are all
compatible on the coax side, i.e., stations connected to the coax via
different types of transceiver can communicate. The incompatibilities
are on the transceiver cable side, and I do not know exactly what all
the differences are. I do know that 2.0 and 802.3 have what is known as
a "heartbeat" feature, which is a self-test of the transceiver's
collision presence circuitry and which puts a signal on the collision
presence pair after each transmission. IEEE 802.3 transceivers use
pin 4 for the shield connection instead of pin 1. There may be
further differences, but these alone may be sufficient to make a
station not function with another version's transceiver. I too
learned this the hard way, since many manufacturers claim IEEE 802.3
compatibility when in fact they are not.
keith gorlen {seismo|umcp-cs}|elsie|cecil|keith
-------------------------------------------------------------------------------
Date: Sat, 3 Nov 84 02:24:45 cst
From: oakhill!ut-sally!ihnp4!oddjob!matt (Matt Crawford)
Phone: (312) 962-8206
Postal-Address: Astronomy & Astrophysics Center 105
5640 S. Ellis ; Chicago, IL 60637
To: oakhill!mot!fred
Subject: Re: Ethernet Transceivers
Organization: University of Chicago, Department of Astronomy and Astrophysics
I had always understood that the difference only affected the
connection between the transceiver and the interface board for
the computer, so that as long as those two components matched,
you're OK. Please reply to correct me if you find that I am
wrong!
Matt University crawford@anl-mcs.arpa
Crawford of Chicago ihnp4!oddjob!matt
-------------------------------------------------------------------------------
Date: Sat, 3 Nov 84 14:57:04 mst
From: oakhill!ut-sally!nbires!mccallum (Doug McCallum)
To: oakhill!mot!fred
Subject: Re: Ethernet Transceivers
What software are you running that requires Ethernet II? If you
are running a fully 802 compatible system (802.3 controller/tranceivers +
802.2 style Link Level Control), then you have to be compatible all the
way around. The biggest problem is in matching the tranceiver with
the controller. An Ethernet I controller will only see errors if attached
via an 802.3 compatible tranceiver. The incompatibility should be fairly
well known. 802.3 added a function to the tranceiver called the "heartbeat".
This heartbeat raises collision for a short time after a successful transmit.
The signal put on the coax should be the same for both Ethernet I and 802.3.
802 changes the Ethernet I type field into a length field. Some 802.3
controllers may reject packets if the length field doesn't have a valid value.
I am currently in the middle of straightening out a mess where our MIS group
installed a corporate LAN and bought 802.3 tranceivers and Ethernet I
controllers. A repeater with mismatched tranceivers doesn't work!
Doug McCallum
NBI, Inc
{ucbvax, ut-sally, allegra}!nbires!mccallum
-------------------------------------------------------------------------------
From: oakhill!ut-sally!ihnp4!ihuxn!jee
Date: 7 Nov 84 07:56:32 CST (Wed)
Subject: Re: Ethernet Transceivers
Could you repost your question with more detailed information, including
brand names. I am not aware of the problem.
1) Trancievers normally implement the physical layer, (i.e. they provide
timing information, read/write packets from/to the network). At that
level (the physical level) there is no difference between IEEE 802.3 and
Ethernet. (But I'm going to check the specs again because it is possible
something like a control lead has changed)
2) In the last year IEEE 802.3 has undergone some small but significant
changes, but they where at the Ethernet link layer. Basicly the
IEEE 802.3 changes reflected a clearer distinction between physical &
link layer. So what does compatiable mean? With IEEE 802.3 not yet
finalized as a standard, which version are you talking about?
It was expected to be accepted as a standard sometime this year,
but the latest version I have for IEEE 802.3 is July 1983.
I suggest you get a copy of both the Ethernet version 2 spec and
the current IEEE 802.3 spec. It is enlightning reading.
John Emrich
AT&T Bell Laboratories
ihnp4!ihuxn!jee
-------------------------------------------------------------------------------