ICMP responses to broadcasts


David A. Bridgham (dab@BORAX.LCS.MIT.EDU)
Sat, 26 Apr 86 14:32:05 est


   Date: 25 Apr 1986 17:03-PST
   From: Steve Deering <deering@su-pescadero.ARPA>

   But the decision to withhold ICMP error reports should be based on
   whether or not the packet was broadcast or multicast AT THE IP LEVEL,
   NOT the local network level.

I quite disagree. If the packet was broadcast at the local net level,
then it should be treated as a broadcast packet for the purposes of
not sending ICMP responses, regardless of what internet address the
packet was sent to. This prevents all manner of nasties which will
occur when someone broadcasts a packet with a non-broadcast IP
address.

   I understand that it can be messy checking for all the different flavours
   of IP broadcast address (at least IP multicast addresses can all be
   recognized with a single test), but these tests tend to be out of the
   performance-critical main path, so I don't think it's such a big deal.

In some systems anyway, the check for broadcastedness *is* in the
critical path. If you only do that check to decide about returning an
ICMP message then it wouldn't be, but gateways, at least, need to
check every packet so that broadcast packets are not forwarded. The
check for a broadcast IP address is also just one test (set the net
and subnet (if used) fields to all 1's and check if the address is all
1's). The problem is that there are many machines that do not use the
IP broadcast address when they broadcast packets. That's what makes
the check for an IP broadcast address slightly messy.
                                                        Dave



This archive was generated by hypermail 2.0b3 on Thu Mar 09 2000 - 14:36:06 GMT