[comp.protocols.tcp-ip] Token Ring Mac Addresses

cire@CISCO.COM (cire|eric) (05/26/89)

I and some of my colleagues have been working with various protocols
running across Ethernet/IEEE 802.3 and Token Ring.  Because of the different
transmission orders we've run into an interesting problem of integrating
networks consisting of these two media.  I'm wondering if others out
there have run into the problem and what if anything has been done.  In
particular I'm interested in any established conventions set by oh DEC
or perhaps Xerox.

On Ethernet and IEEE 802.3 media data is transmitted MSByte to LSByte but
the first bit transmitted is the least significant.  On Token Ring
media the bytes are shipped out in the same order but the transmission
starts with the most significant bit.  In either case the very first
bit transmitted is the Individual/Group (multicast) bit and the next
bit is the Universal/Local address administration bit.

If for example I'm working with DECNet I would have the Ethernet address
AA00.0400.06CC assigned to one of my nodes.  If this address were used
on a Token Ring it would be multicast.  One possibility is to swap each
byte.  Thus the above address becomes 5500.2000.6033.  Perhaps that is why
AA was chosen.  Similar problems present themselves for XNS.

Any thoughts will be appreciated.  Please mail directly to me and I will
summarize to the net.  Please do not respond directly to this list.  It
already deal with a lot of traffic.

-c
cire|eric

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

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

tmallory@BBN.COM (06/07/89)

Eric,

I've been a little behind in my mail, and have saved your message for a later
moment, which is now.  Though I don't have a copy in front of me, my memory is
that the cover letter which the IEEE sent with a block address assignment, as
well as the address itself, explicitly addressed the fact that the bit orders
for 802.3 and 802.4-5 are reversed.  The assigned address block refers to the
bits as they appear on the wire(fiber, etc.).  Our address was printed in
three forms:  binary network order, 802.3 hex, and 802.5 hex.  Reversing the
bits in each byte is the correct way to go for the different network types.

BTW There was a recent tip-ip posting concerning the hardware addresses from
some well-known manufacturer's equipment which did not match the assigned
address block(perhaps AT&T?), that appeared to be a bit-reversal problem.

Hope this is helpful, and that you didn't get too many versions of this
message.

Tracy Mallory