Yakim Martillo (bloom-beacon!martillo@ucbvax.Berkeley.EDU)
15 Aug 87 23:41:18 GMT
In article <8708131534.AA14351@ucbvax.Berkeley.EDU> cain@EDN-UNIX.ARPA (Edward A. Cain) writes:
>Is there a commercial product (gateway) which supports ethernet multicast?
>Needs to take an IP datagram from Arpanet, recongize that the IP address
>is really a group address, and multicast it onto an ethernet.
I believe that for hosts directly connected to an IMP via 1822 IP
addresses uniquely identified the IMP interface to which a host is
connected. The unique mapping of hardware interface to IP address is
quite natural in this case. When TCP/IP moved onto ethernet, the
mapping of IP address to physical interface was maintained though for
a multihomed host with a separated IP address on each network (of any
type of communications subnet) there should be no problem with
handling an IP packet on any interface with any IP address as long as
IP processing is done in the host or as long as the communications
controllers can communicate with one another over the backplane as is
possible with DEC machines. Also when people began putting PCs onto
ethernets, dynamically assigning IP addresses to PCs seems to be quite
reasonable. Further when ISDN interfaces become available,
dynamically assigning IP addresses to sets of switched physical or
virtual circuits will probably make sense (especially for certain
cases of security). Thus my questions are:
Is there such a thing as an IP *group* address and if
there is what is it used for?
Next if there are IP group addresses how do they work
in an internetting environment.
If there are IP group addresses, does it make any sense to
map them to ethernet multicast addresses (which I guess would
be assigned to groups of hosts dynamically) and if it does how
would one go about handling multinetwork IP group addresses
for which multicasting would be clearly insufficient.
This archive was generated by hypermail 2.0b3 on Thu Mar 09 2000 - 14:38:49 GMT