> > Globally unique. The point here is that the Ethernet standards require > > a globally unique MAC address *per box*, not necessarily per interface. > > The "Sun way" of storing a MAC address in EEPROM and configuring all > > network cards with the same MAC address was perfectly compliant, just > > somewhat unusual. > > But that's not the whole story, right? > > The only way this can work is if one assumes that each interface of that box > is connected to a different IP subnet. I don't even think Ethernet rules are > what matters here. If the box should happen to be dual homed to the same IP > subnet, that Sun scheme doesn't work. (I suppose unless only one of the box's > interface is active at any given time.
Obviously. And I believe more modern Sun boxes actually have a MAC address per interface. All I'm saying is that there is nothing wrong, standards-wise, in having *one* globally unique MAC address per box. Having the same MAC address shared between two (or more) boxes *is* wrong, standards-wise. And unfortunately that's exactly what we're seeing in practice. Not often, but "often enough" that it may not be safe to ignore the possibility. Steinar Haug, Nethelp consulting, sth...@nethelp.no -------------------------------------------------------------------- IETF IPv6 working group mailing list ipv6@ietf.org Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6 --------------------------------------------------------------------