On 2/20/19 12:23 PM, Paul Koning via cctalk wrote:
Please note that among LANs, there is Token Ring (802.5) and there is
everything else.
I think it really depends on how you look at them.
From a frame formatting point of view, Ethernet is the odd ball when
looking at how TCP/IP is carried.
Everything other than Ethernet (802.3) uses 802.2 or a medium specific
varient of 802.2. Then there's Ethernet which predominantly uses either
Ethernet II for TCP/IP or 802.3 (a.k.a. "Raw") Ethernet frames for IPX.
FDDI is like Ethernet and like 802.4. Token Ring is the oddball because
(a) it doesn't have proper multicast addresses, and (b) for some reason
IBM invented source-routed bridging and tied that to Token Ring.
Does it actually need a broadcast address like Ethernet when the ring
passes through all the stations? Or is that functionally comparable to
a multicast?
FDDI is in no way at all like Token Ring. The only thing the two have
in common is "token" and "ring". The MAC protocol is utterly different;
the closest relative is 802.4 Token Bus. And as far as addressing is
concerned, FDDI is like 802.4 and Ethernet, with real multicast and
general use of normal transparent bridges.
The only complication with FDDI (and 802.4, if you could find it)
is that it only has 802.2 frames, not classic-Ethernet (with 16 bit
protocol types). So an FDDI to Ethernet bridge has to translate Ethernet
frames to an 802.2 based encapsulation. That is done by converting them
to SNAP frames, as described in RFC 1042.
Intriguing.
$ReadingList++
Bridges like the DECbridge 500 and DECbridge 900 will do that; I assume
Cisco does likewise.
FDDI didn't live all that long because 100 Mb Ethernet replaced it, but
while it was out there it made a fine backbone for Ethernet-based LANs.
:-)
--
Grant. . . .
unix || die
