[comp.protocols.tcp-ip] DecNet on TR

cire@CISCO.COM (cire|eric) (08/10/89)

I know this is perhaps blasphemous so I'll put on my asbestos suit.

As far as I know the only media that DecNet currently runs over is
serial lines and ethernet.  Are there other media that DecNet is being
implemented on?  In particular are there other implementations of
DecNet on Token Ring?  I know that Sun has a Token Ring interface and
Apple is working on one.  Both of these folks have some form of 
potential relationship with DecNet.  So there is some likelihood that
it is being worked on.

Cisco has a DecNet on Token Ring implementation and we would certainly 
like to have it play with the other kids on the block.  I've run into
a number of interesting little details that would need to be agreed on
to accomplish this.

1) Ethernet DecNet interfaces set their hardware address to
   AA00.0400.xxxx (where xxxx is determined by area.node).  This can
   not be used by Token Ring interfaces because it is a Token Ring
   non-specific address.  One possibility is to bit swap all the bytes
   of the address.  This is the recommended practice given by IEEE.
   So the address becomes 5500.2000.yyyy.  Note that AA on ether and
   55 on Token Ring are both Locally Administered addresses.  This is
   important.

   The choice isn't particularly important as long as it is agreed on.
   One other significant detail on this choice is whether we want this
   stuff to operate reasonably over transparent Token Ring-Ethernet
   bridges.  If so then the bit swap is probably the way to go.  I say
   probably because there are significant technical details that would
   need to be solved for such an thing to exist.  Primarily because of
   the bit ordering problems.  These are significant.  I'd recommend that
   we don't worry about Ether to Token Ring transparent bridging.

2) DecNet on Ethernet also uses two multicast address for talking to
   all Nodes and all Routers on this cable.  All Token Ring hardware
   I've seen and all hardware based either on the IBM chipset or TI
   chipset only support a single multicast address (called the Group
   address, not to be confused with the Group bit).  Common practice
   I've seen on Token Ring so far is to use what is called a Functional
   address.  A Functional address is a bit significant address.  Two
   bits would have to be chosen that don't conflict with other uses
   of the Functional addresses.  Currently the following are the ones
   that seem to have gained some solidity:

   NOVELL		C000.0080.0000
   CLNS ESIS all IS	C000.0010.0000
   CLNS ESIS all ES	C000.0008.0000
   all Bridges		C000.0000.0100  **
   NETBIOS		C000.0000.0080
   Config Rpt. Server	C000.0000.0010	* a.k.a. Lan Manager
   Ring Error Monitor	C000.0000.0008	*
   Ring Parameter Srv	C000.0000.0002	*
   Active Monitor	C000.0000.0001	*

	*  = from standard IEEE 802.5
	** = from draft standard IEEE 802.5D (for review only)


   Note that there isn't as far as I know any central authority that
   has responsibility for these bits other than some of the lower ones.
   Those have been set by IEEE and earlier IBM.  The other bits appear
   to be up for grabs.  Like the address in point 1 above, which two bits
   chosen need to be agreed on for interoperability.


Comments?  Also if there is a better place to post this kind of message
please feel free to inform me.  I would say this list has the largest
readership of folks interested in interoperability issues that I know
of.

thanks for you time,

-c
cire|eric

Eric B. Decker
Token Ring Development
cisco Systems - engineering
Menlo Park, California

email:	cire@cisco.com
uSnail: 1360 Willow Rd.,  Menlo Park, CA  94025
Phone : (415) 326-1941

adelman@TGV.COM (Kenneth Adelman) (08/11/89)

> I know this is perhaps blasphemous so I'll put on my asbestos suit.

    At least my answer will be less blasphemous...

> As far as I know the only media that DecNet currently runs over is
> serial lines and ethernet.  Are there other media that DecNet is being
> implemented on?

    Yes.  As part of our MultiNet (VMS TCP/IP) product we encapsulate
DECnet packets in IP datagrams.  This lets you run DECnet over the
wide variety of devices supported by IP, and across wide-area networks
which only support IP.

    At SRI we even run DECnet over IP over an ethernet because we have
Interlan cards in our 750 and 780 and by layering DECnet over IP we
don't need to buy DEC ethernet cards (for you non-DECies, DECnet/VAX
only supports DEC's own ethernet cards).

						    Kenneth Adelman
						    TGV

bob@giza.cis.ohio-state.edu (Bob Sutterfield) (08/14/89)

In article <8908100159.AA18417@ucbvax.Berk?ley.EDU> cire@CISCO.COM (cire|eric) writes:
   As far as I know the only media that DecNet currently runs over is
   serial lines and ethernet.

In days of yore DECnet's only medium was a parallel interface like the
Unibus DR11-W.  Ethernet was a major step forward at the time, and its
advent was the cause of much rejoicing.  It probably still runs on
parallel interfaces.  Also, you might consider the fiber links to the
DEC mass storage server (sorry, I've forgotten the name/number) to be
a token ring, and I believe they can be a host's only connections to
the rest of a DECnet.

   Are there other media that DecNet is being implemented on?  In
   particular are there other implementations of DecNet on Token Ring?

Proteon's networks can pass DECnet, much as they can pass IP and XNS.
OSU has some PROnet-80 routers configured for DECnet service, though
throughput suffers markedly if DECnet shuffling is enabled.  It's
something about DECnet relying on {broad,multi}cast and that absorbing
too much juice from the router's CPU.  I don't recall the details; we
were just happy when the DECnet folks got their own router and freed
up our box for just IP and XNS.

bzs@ENCORE.COM (Barry Shein) (08/17/89)

Decnet also runs over DMR11's (or it's older version, the DMC11.)

It had a few modes of interface including serial. At BU we ran DECNET
over DMR's using triax between a 2060 a VMS/780. We ran the DMR's at
250Kb, it ran at up to 1Mb but I remember field service mumbling
something about the DN20 or some such at that speed so we turned it
down.

	-Barry Shein

Software Tool & Die, Purveyors to the Trade
1330 Beacon Street, Brookline, MA 02146, (617) 739-0202