JJ, This argument comes up frequently from people that are looking at IPv6 for the first time. First, ignore the number of bits for the moment and look at the differences between the versions. In IPv4, your ISP allocates a host address, while in IPv6 the allocation is a subnet at a minimum, and should be a block of subnets in the general case. Since the dial allocation is a subnet, there is no requirement to use NAT, when a simple layer-2 bridge will do the job. There is also a concept frequently referred to as multi-link subnets, which differs from a layer-2 bridge only in the fact that forwards based on layer-3 packets rather than layer-2 frames, but considers all interfaces to be part of the same logical layer-3 subnet. Using either of these technologies you will be able to have a substantial number of nodes connected through a dial link.
Now as for the number of bits, there are several capabilities that have been built based on the availability of 64 bits for the interface id. Auto-configuration based on EUI64/48 mac is just the first. Privacy (RFC 3041) requires a large number of bits to avoid collisions; automated allocation of multicast addresses (draft-ietf-ipngwg-uni-based-mcast-03.txt) needs space to divide the group space by prefix; and ISATAP (draft-ietf-ngtrans-isatap-02.txt) uses the space to encode a tunnel endpoint for non-multicast IPv4 infrastructures. Others may show up to solve specific problems, so trying to be too conservative with bits is actually wasteful in terms of time and flexibility of deployments. I hope this helps, Tony > -----Original Message----- > From: JJ Behrens [mailto:[EMAIL PROTECTED]] > Sent: Wednesday, January 23, 2002 11:18 AM > To: Tony Hain > Cc: Michel Py; Keith Moore; Irina Dayal; [EMAIL PROTECTED] > Subject: RE: IPv6 Addr/Prefix clarification > > > > From: Tony Hain <[EMAIL PROTECTED]> > > Keith is technically correct, a site may choose to subnet the low 64 > > bits. It might be unwise to do so, because they would be > giving up the > > autoconfiguration & privacy capabilities, but that is their > choice. They > > may also find acquiring routing hardware that works beyond > the 64 bit > > prefix to be a challenge ... > > Please forgive me for being a newbie, but it seems wise to allow > subnetting of the lower 64 bits. Afterall, it would be terrible if my > dialup ISP assigned a /64 to me, and I had to rely on some > IPv6 mythical > NAT to do subnetting! > > A /64 is really quite a lot of address space to subnet; without the > ability to subnet, a /64 seems wasteful. In response to the address > autoconfiguration using MAC addresses argument, address > autoconfiguration > can be done using something other than the MAC address in those > circumstances. The fact that neighbor solicitations are used > to verify the > uniqueness of tentative addresses is proof enough that MAC > addresses were > never meant to be the end all and be all of address autoconfiguration. > > -jj > -------------------------------------------------------------------- IETF IPng Working Group Mailing List IPng Home Page: http://playground.sun.com/ipng FTP archive: ftp://playground.sun.com/pub/ipng Direct all administrative requests to [EMAIL PROTECTED] --------------------------------------------------------------------
