bruce@ssc-vax.UUCP (Bruce Stock) (11/13/86)
I have always been under the assumption that Ethernet Physical Addresses were virtually "cast in concrete" , ie unchangable. I had heard that Xerox maintains a registry of physical addresses, and that manufacturers may reserve blocks of addresses for their products, thus avoiding any address conflicts. However recently I have come to understand that DECNET, for example, takes a DEUNA with a physical address something like 08-00-2b-03-fc-83, and replaces it with a physical address like aa-00-04-00-25-04, where the "25" corresponds to the DECNET node number. Digging a bit further, I note that our Interlan ethernet boards also have the capability for over-riding the built-in physical address. Would any one care to shed some light on why this is necessary or desirable? Does it not tend to increase the risk of an address conflict on a large, multi-protocol, mult-vendor net? Thanks, Bruce E. Stock uw-beaver!ssc-vax!bruce
rpw3@amdcad.UUCP (Rob Warnock) (11/16/86)
+--------------- | Digging a bit further, I note that our Interlan ethernet boards also have the | capability for over-riding the built-in physical address. | | Would any one care to shed some light on why this is necessary or desirable? | Does it not tend to increase the risk of an address conflict on a large, | multi-protocol, mult-vendor net? +--------------- Note that an Xerox XNS node is required to have the *SAME* Ethernet address on *ALL* nets that it is connected to. (In XNS the Ethernet address *is* the host address. In TCP/IP language, the physical address *is* the IP address). Since the two (or more) controllers used in that situation will come with more than one address, all but one of them has to change. Typically, you simply pick the largest (or smallest) address of any of the controllers, and set that address into all of them. There is no conflict, since you own all of those addresses anyway. Rob Warnock Systems Architecture Consultant UUCP: {amdcad,fortune,sun}!redwood!rpw3 DDD: (415)572-2607 USPS: 627 26th Ave, San Mateo, CA 94403
normt@ihlpa.UUCP (N. R Tiedemann) (11/18/86)
> > I have always been under the assumption that Ethernet Physical Addresses > were virtually "cast in concrete" , ie unchangable. I had heard that > Xerox maintains a registry of physical addresses, and that manufacturers may > ... > replaces it with a physical address like aa-00-04-00-25-04, where the "25" > Bruce E. Stock > uw-beaver!ssc-vax!bruce Xerox does keep a registry (soon to be taken over by IEEE) of "all" ethernet physical addresses. The thing to not about this address change is the "aa" in the first byte. Normally the first 3 bytes are the manufacture's code and the last 3 are the device. Now, everyone knows that the first bit (LSB) of the first byte specifies a group address (vs. a physical address), but what is not know by many is the second bit is a local/global administered bit. If that bit is set, the device is not registered with Xerox, and there is a possiblity of two devices having the same physical address. This is normally used in closed systems, where outside devices are not attached. (Restricted LAN). So once the network is set-up the individual devices can "change" their physical addresses, and set this bit and be "legal". Most often this is used where a network is initialized and not changed, i.e. no new nodes added or dropped. We use this when we use "ethernet" to communicate between processors in a multiprocessor system. No one else is supposed to be connected to this network, it doesn't go outside of a couple of frames of equipment, and we may build lots of them, but they are NEVER interconnected. Norm Tiedemann ihnp4!ihlpa!normt AT&T Bell Labs "For a good time, read an IEEE standards spec."