On Oct 10, 2013, at 8:44 AM, Brian Hechinger <wonko at 4amlunch.net> wrote: On Thu, Oct 10, 2013 at 02:36:04PM +0200, Peter Lothberg wrote: FDDI/CDDI is a dual ring token ring bus, with 4470 MTU byte packets, it has 802.-- frames. DEC had a mode where you turned the token off and used it for ptp full duplex. I didn't know about the ptp thing. That's nifty. For example cisco/cabletron/crecendo had ethnernet switches with a FDDI uplink, that you could use. DEC made one as well, it was that large modular thingie. I used to have one. Never got it powered on as it was enormous. But you need nothing to build a FDDI ring, its a A and a B ring, you can just plug the cards together with fiber-patch-cables. Unless you have one of those obnoxious single attached station cards. -brian Ah, time to dust off some dormant memories. I used to work on the FDDI standard at DEC; this stuff is familiar. "CDDI" is marketing slang; it is not standard terminology. FDDI is different from Ethernet; the MAC layer protocol is completely unrelated. It's quite similar to token bus (802.4), actually. (The only thing it has in commoni with 802.5 is the words "token" and "ring" -- apart from that, the two protocols operate completely differently.) FDDI connections have a "type", which can be "A", "B", "S", or "M". "M" ports exist on concentrators. NICs will have A and B ports, if there are two connectors on the NIC ("dual attached station" or DAS) or an S port, if there is one connector (single attached station or SAS). You have a number of topology options. If you have DAS NICs, you can wire any number of them together in a "dual ring". That's the original FDDI topology, before DEC forced concentrators to be added into the standard. To do that, connect the NICs in circular fashion, A to B. Connected that way, loss of any single connection is handled transparently. If you have SAS NICs. you can connect a pair of them (S to S). If you have DAS NICs plus or or two SAS, wire the DAS NICs A-B in a chain (essentially a dual ring cut open). Then connect a SAS to each end (or just to one end, if you have one SAS). There is no redundancy in this config. Finally, if you have any concentrators, you can build a tree config out of those. If so, the M connectors connect to the NIC connectors (A, B, or S), and the concentrator's A and B connectors either connect to M ports higher up in the hierarchy, or in a dual ring if you have a ring of concentrators, or nowhere if you're at the root of a tree. FDDI fiber connectors are standardized but different from fiber connectors used by other networks. The connector is fairly large, flat rectangular with a shroud covering the fiber ends. Connectors are keyed to match the port type, though you can omit the keys and just wire carefully. Standard fiber is 62.5/125 micrometer, but 50/125 also works. You can find more in the DEC Technical Journal, Vol. 3 No. 2, spring 1991. Or the relevant ANSI/ISO standards if you are a masochist. The topology rules are described fairly well in the concentrator article in DTJ, but their full glory can be found in the FDDI "SMT" (station management" standard. paul