On Sunday 10 April 2005 19:42, Pierre THIERRY wrote: > Scribit Jeroen Massar dies 30/03/2005 hora 17:55: > > Scale the above up too two networks with 10.000 devices each, have fun > > renumbering then.... > > Is this the very argument to deprecate site-local? Renumbering is, > AFAIK, a totally automatic process, if all addresses are automatically > configured with router advertisement. > That renumbers the IP addresses on the machines. Now you've got to deal with DNS servers, and anything with a static IPv6. Assuming you have to renumber a DNS server, you then have to go through and handle ensuring that everything has the new DNS server address. On top of that, you're assuming that all administrators involved are prepared to renumber to accommodate each other's site-local addresses. Imagine a scenario involving two businesses; each business has used site-local addresses until they got themselves a global IPv6 prefix. Due to the pressures of work, neither business has completed renumbering from site-local to global addresses (the network works, so who cares if it takes an extra 6 months to complete the renumbering). An employee from business 1 makes a VPN connection from business 2 back to his home network. A service he needs to access hasn't yet been renumbered, and its old address conflicts with one on business 2's network. How does the laptop know that fec0::/16 addresses from a DNS server on interface vpn0 should be talked to via vpn0, while a 2001::/16 address from the same DNS server should be talked to via eth0? Now throw in access to the home site via UMTS, so that there's three interfaces with fec0::/16 addresses, two of which can be used to connect to one set of fec0::/16 addresses, and one of which has a different, numerically identically set of fec0::/16 addresses. Yes, it would be possible to build in some sort of conflict resolution to handle this, but surely it's simpler to use up a small portion of the IPv6 address space for unique local addresses, thus ensuring that local addresses are unique on each site? -- Simon Farnsworth
Attachment:
pgpRFZnkw2a1g.pgp
Description: PGP signature