heyas!

[mikebw at bilow.bilow.uu.ids.net (Mike Bilow)]

Peter Breton wrote in a message to Mike Bilow:

 PB> 3) You *must* use the same Ethernet card, because they check
 PB> the hardware address of the card.

The ROM address of the Ethernet card is not directly accessible from the
network, but is simply read by software.  Obviously, most software has no
reason to use an address on the wire different from that read out of ROM, but
nearly all software has this capability.  Even standard NDIS or ODI stacks
allow manually overriding the Ethernet hardware address in PROTOCOL.INI or
NET.CFG, in case you find yourself in a situation where this is necessary.

Longshine Technology, for example, was notorious for selling budget price
Ethernet cards which all happened to have the same hardware address burned into
ROM.  Their trick was to send out a free sample, which would obviously work
perfectly, and then sell a bunch to the unsuspecting customer.  Diamond Flower
International apparently did not want to pay a registration fee for a block of
Ethernet hardware addresses, so they made all their cards use the ASCII
characters "DFI" as the first three of the six address bytes.  It was stunts
like this which led to having software support for manual overrides.

HP/UX has a scheme where they use multiple Ethernet cards on the wire with the
same address in different machines, providing live fail-over instantaneously
and transparently to the network.  If this was not done, then significant delay
would be introduced into the transition; an ARP cache timeout applicable to an
IP network defaults to 15 minutes.

Linux/Unix should provide for manual control of an Ethernet hardware address
through the "ifconfig" facility.  You should be able to set this to anything
you want, assuming you are not trying to boot from the network using BOOTP,
DHCP, or some similar special protocol largely ahead of the operating system.
 
-- Mike