The Routing Over Large Clouds Mailing List Archive by date

[Dave Katz <dkatz@cisco.com>]

   The quote from me above means:

	   ar$hrd = X, media is Ethernet, address is 48 bit MAC.
	   ar$hrd = Y, media is FDDI, address is 48 bit MAC.
   (and of course)
	   ar$hrd = Z, media is ATM, address is either E.164 or NSAPA.

This is *exactly* why the hardware type was broken in ARP.  There were
a bazillion IP-over-Ethernet boxes in the field that expected type code
1, and could not talk to anything else.  Along came FDDI;  1103 naively
used the same type code as "802".  Result--you couldn't interoperate
because you could not fix the deployed base.

   What's wrong there? I never said that a different ar$hrd value
   has to mean a different address type from any other ar$hrd value.
   I dont know why anyone in their right minds would assume it had
   to.

The problem is the opposite side--how does a box that likes media type X
know that media type Y is the *same* address type, particularly if media
type Y wasn't invented when the first box was deployed?