oberman@LLL-ICDC.ARPA ("Oberman, Kevin") (05/29/87)
Marc Shannon commented about the order of starting Ethernet software as follows: >The only trick we had was that DECnet must be started before TCP/IP...It >only wants to speak to certain network devices, I guess. (Sounds like the >making of a computer soap opera!) Perhaps a bit of an explanation as to why would make things clearer. When DECnet starts running over Ethernet it does something a bit unusual. It resets the unit's Ethernet address to the form AA-00-04-00-ll-hh. `ll-hh' is the DECnet address of the node with the bytes swapped. For example a DECnet node with an address of 10.1 would set its Ethernet address to AA-00-04-00-01-28. Swap the last two bytes and you get 2801. In decimal this is 10241 (10*1025)+1. 3.106 would be AA-00-04-00-6A-0C. C6A(16) is 3178(10) or (3*1024)+106. This makes it very easy to know what DECnet node is associated with what Ethernet address. But it requires DECnet be started before any other protocol. This includes both thirs party software and DECs own. You can't start LAT and then start DECnet. DECnet will move the Ethernet address right out from under LAT (or whatever). This presents a problem if you want to use some other software product which changes the physical Ethernet address of a node. No chance. I really question if the system DEC uses is a good idea, but its there for now. I assume that Phase V will have to change this because the larger DECnet address size (ISO format) would be bigger than the Ethernet address could hold. R. Kevin Oberman Lawrence Livermore National Laboratory arpa: oberman@lll-icdc.arpa (415) 422-6955 Disclaimer: I an not endorsing any product. I work for Uncle Sam and to do so would make him upset. I further take no resposibility for the accuracy of this information. I believe it is correct, but if it's not I can only say "Sorry". ------