> DHCP works on non-local LANs because of special forwarding
 > performed for DHCP/BOOTP by routers.

I'm not sure why you call this "special forwarding" -- it's simply normal
DHCP relay agent behavior.  Also, there is no requirement that a DHCP
relay agent be an IP router.

 > I'd recommend updating the spec to remove the reference to
 > "inducing broadcast storms" - this has nothing to do with why
 > it should be consolidation private, if in fact it should be private.

I think it's relevant.  If there's no use-case, we should not provide rope
to hang oneself.

 > Given that we have IP_TTL as an option for getsockopt,
 > why shouldn't we also implement that for setsockopt?

We do have it for setsockopt().  It affects unicast traffic.

 > How would IP_BROADCAST_TTL interact with IP_TTL?

It doesn't interact since IP_TTL only affects unicast traffic.

 > To look into the future, can you explain why we would need
 > to support setting IP_TTL, IP_BROADCAST_TTL and
 > IP_MULTICAST_TTL, all on the same socket?

One affects unicast, one affects broadcast, and one affects multicast.
IP_MULTICAST_TTL is the only one that's a stable API.  BTW, this same
split exists with IPv6: IPV6_UNICAST_HOPS and IPV6_MULTICAST_HOPS (of
course there is no V6 broadcast); both are documented in ip6(7P).

-- 
meem

Reply via email to