JLarson.pa@XEROX.COM (12/13/88)
Re: What was the rationale for 48-bit Ethernet addresses? They are never used beyond the "Local" Area Network; Machines may move from one "Local" Area Network to another (and even between non-local networks). Unique 48-bit Ethernet addresses eliminates a layer of a network administration that would be required otherwise. Also, unique host numbers are useful in large distributed systems (for generating unique identifiers). It just happens to be a nice optimization that a unique 48-bit XNS host number and a 48-bit Ethernet address are the same. There is some danger though for a distributed system algorithm which absolutely depend son a unique ID. We have seen cases in the Xerox Internet where 48-bit host numbers were not unique as intended. (Apparently PROM burners can get stuck on the same number once in a while.) John Larson Xerox PARC
stev@VAX.FTP.COM (12/14/88)
*Re: What was the rationale for 48-bit Ethernet addresses? They are never * used beyond the "Local" Area Network; * *Also, unique host numbers are useful in large distributed systems (for *generating unique identifiers). It just happens to be a nice optimization *that a unique 48-bit XNS host number and a 48-bit Ethernet address are the *same. * *There is some danger though for a distributed system algorithm which *absolutely depend son a unique ID. We have seen cases in the Xerox *Internet where 48-bit host numbers were not unique as intended. *(Apparently PROM burners can get stuck on the same number once in a while.) * *John Larson *Xerox PARC * part of the win also was that the numbering was "distributed" have the ethernet address was vendor code, half was "serial number". (i realize that "serial number" is not the real title, but it always suprised me that most boards has a diffrent serial number if one is mentioned.) makes it easier to figure out the "class" of machines that is troubling you, and gives them the freedom to do with their numbers as they wish. stev knowles stev@ftp.com ftp software
henry@utzoo.uucp (Henry Spencer) (12/16/88)
In article <881212-182513-5931@Xerox> JLarson.pa@XEROX.COM writes: >There is some danger though for a distributed system algorithm which >absolutely depend son a unique ID. We have seen cases in the Xerox >Internet where 48-bit host numbers were not unique as intended. >(Apparently PROM burners can get stuck on the same number once in a while.) Also, production-line people are not used to the idea that each system must be *different* -- they are very much organized around exact duplication of hardware, PROMs included. There are tales of this causing trouble at several companies making Ethernet gear. -- "God willing, we will return." | Henry Spencer at U of Toronto Zoology -Eugene Cernan, the Moon, 1972 | uunet!attcan!utzoo!henry henry@zoo.toronto.edu