Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.3 alpha 4/15/85; site ubvax.UUCP
Path: utzoo!linus!decvax!decwrl!amdcad!cae780!ubvax!skip
From: skip@ubvax.UUCP (Skip Addison Jr)
Newsgroups: net.lan
Subject: Re: Repeaters, Bridges, Routers, and Gateways
Message-ID: <442@ubvax.UUCP>
Date: Mon, 17-Feb-86 14:01:48 EST
Article-I.D.: ubvax.442
Posted: Mon Feb 17 14:01:48 1986
Date-Received: Wed, 19-Feb-86 07:33:38 EST
References: <1174@decwrl.DEC.COM>
Reply-To: skip@ubvax.UUCP (Skip Addison Jr)
Organization: Ungermann-Bass, Inc., Santa Clara, Ca.
Lines: 108
Summary: lan vendor survey/poll

In article <1174@decwrl.DEC.COM> gassman@jedi.DEC writes:
>
>
>WHOA... it's time to have a few definitions standardized!
>
In article something or another, Skip Addison @ Ungermann-Bass writes:
>>Bridges get involved at the network layer.  IEEE 802.3 & 802.4 only
>>specify up through the data link layer.  What bridge you need depends
>>on what network layer protocols you need.  For example, Ungermann-Bass
>>provides networks based on a variety of different protocols.  Which
>>bridge software you get depends on which network protocols you're using.
>>Like Ungermann-Bass's, Bridge's and DEC's bridges will only work on
>>networks that use their protocols.  Fortunately, we've all started basing
>>our products on standard protocols, so depending on your system,
>>vendor A's network interface units may communicate using vendor B's
>>bridges.
>> .
>> .
>> .
>>-- Skip Addison
>>   {lll-crg, decwrl, ihnp4}!amdcad!cae780!ubvax!skip
>>   Ungermann-Bass, Inc
>>   (408) 496-0111
>
>
>Skip Addison of Ungermann-Bass is greatly mistaken about
>what DIGITAL's LANbridge 100 will do.  He states that DEC's
>bridge works at the OSI Network layer, and is therefore protocol 
>dependent.  THIS IS NOT THE CASE!!!!  BRIDGES work at the OSI 
>datalink layer, and therefore MUST BE PROTOCOL INDEPENDENT.
>DEC's LANbridge 100 is indeed so.  What Ungermann-Bass sells as a 
>BRIDGE is actually a ROUTER, a router being a protocol DEPENDENT 
>device working at the network layer.  
> ...
>DIGITAL's bridge forwards to the 'B' LAN, *ANY* packet who's 
>destination station is not on the 'A' LAN.  There also is optional 
>bridge management software to do protocol or multicast filtering.  
>Two models exist.  One with two ethernet plugs, and one using up to
>2000 meters of fiber optics.  Sorry for the commercial, but putting 
>out misinformation about LAN products is not helping the standardization 
>cause at all.  REPEATERS, BRIDGES, ROUTERS, and GATEWAYS are words
>that vendors and users must all agree on, while we trek towards OSI.
>Skip, please contact me about how DEC and UB can start to agree on
>what we call standard products!  
>
>Bill Gassman
>Networks and Communications
>Digital Equipment Corp.
>(603) 884-0192

The original posting referred to DEC's bridge and I assumed (yes, I know
what assume breaks down as :-) that DEC's bridge did the same thing as
everything else I've heard referred to as a bridge.  Incidentally, I learned
my network terminology long before I began working for U-B.

Rather than saying I'm right and you're wrong or some other sort of non-
helpful posting, let's take a vote of people who have worked in the LAN
field more than a year or two.  I'll post my definition of the terms
(which I assume (there's that word again) is close the U-B party line, but
not necessarily exactly) and if you (general you, not you Bill) agree with 
that you should mail me a message saying so.  If you disagree on some 
substantive portion of the posting, post the differences.

Repeater -- Performs a translation from one physical cable segment to another.
The important part is that it does not change the immediate destination or
source addresses.  To do so would be a routing function.  Generally introduces
something like a one bit-time delay, but there are exceptions as noted below.
Also generally reproduces jamming signals, fragmented frames, etc.

Bridge -- One case of an internetwork router.  Changes the immediate 
destination as necessary to further the purpose of getting the packet to its
ultimate destination listed in the Network Layer header.  Changes the
immediate source to its own address.  Of course, if it determines the packet
does not need to be forwarded, it does nothing with it.  A bridge performs
a store and forward operation.  Each 'side' of the bridge can be considered
a seperate signal space.  Jamming signals, frame fragments, etc are not
re-transmitted.  

Gateway -- Another case of an internetwork router.  In addition to the
functions of a bridge, it provides protocol translation at some higher
layer.  For example, something that translated TCP/IP to Xerox's Internetwork
Datagram Protocol and Sequenced Packet Protocol would be considered a
gateway.  Translating an IBM 3270 data stream to VT100 escape sequences
might also be called a gateway.

With the advent of so many network protocols operating on top of Ethernet,
many customers have found it helpful to have what I'll call an an Ethernet 
'linker' for the moment.  It operates just on the information that's available
in the official Ethernet header.  It does not change anything in the Ethernet
header.  Ungermann-Bass calls that a Buffered Repeater, but I don't believe 
that that term is really standardized.  What you call one of these 'linkers' 
is pretty much up in the air.  These linkers are not smart enough to serve 
in a complex network topology with many signal spaces, loops or cycles, 
parrallel load sharing 'linkers' or whatever.  But due to their Network Layer 
protocol independence, they do have some use.


We have quite a few 'experts' on the net.  Also a few real experts.  Some have
even written books on the subject.  Care to comment Andy?  I'd also be
interested in hearing from other network vendors.

Bill, your views of the layers, etc are probably pretty well represented
above in your followup to my previous article.  Is there anything missing?


-- Skip Addison
   {lll-crg, decwrl, ihnp4}!amdcad!cae780!ubvax!skip
   (408) 496-0111