On Friday 06 October 2006 03:13, Hans-Werner Hilse <[EMAIL PROTECTED]> wrote about 'Re: [gentoo-user] OT - ipkungfu not': > On Thu, 5 Oct 2006 18:53:55 -0500 "Boyd Stephen Smith Jr." > <[EMAIL PROTECTED]> wrote: > > > So it would be sufficient to > > > specify a /29 netmask (255.255.255.248). > > However, we can't specify a /30 because two addresses in each block > > (the highest and the lowest) are reserved for "network" (anycast) > > and "broadcast" (multicast). > While this is correct when going for the standard common > implementation, linux will happily accept a broadcast address _outside_ > of the specified network
Yeah. I'm not sure why. It makes my little brain hurt just thinking of it. But, broadcast is not used much. More than anycast, sure, but, not much. That said, a router that did only understood standard broadcast [1] will send those packets to every known machine with the correct netmask. Thus, that address is reserved and should not be used unless you really know your setup. > And anycast is > mainly a routing issue and AFAIK not even implemented in linux. Anycast is virtually unused anywhere. I'd imagine it could be used in some crazy layer 3 clustering solution, but I've never actually seen it used. That said, sending a packet out to the anycast address is dangerous. A router that did implement anycast [1] need not send out those packets to the machine you believe you've assigned that address to (it may route it to any known machine with the correct netmask). Thus, that address is reserved and should not be used unless you really know your setup. Linux is nice and does let you assign this address, for a number of reasons. > That's why I wrote it is likely to break > routing and broadcasting. Using the network or broadcast addresses as an assigned address is likely to break the routing, but using a netmask that is of an "unusual" length 30, 29, or 28 (as opposed "usual" lengths of 8, 16, or 24 *only*) will not, as long as the computers all agree on a netmask -- which is required even for "usual" length netmasks. -- "If there's one thing we've established over the years, it's that the vast majority of our users don't have the slightest clue what's best for them in terms of package stability." -- Gentoo Developer Ciaran McCreesh [1] ...and knew your netmask. It's not transmitted, and in these days where we use CIDR it can't be determined from the IP. (I believe classful networks were assigned ranges, but I'm not sure.)
pgpb8bGenqQvI.pgp
Description: PGP signature