James Kempf wrote:
Correct. However, the v6 addressing spec prohibits the use of an
anycast address from being used as the source address in a datagram, or
being bound to an interface on a host. These two restrictions
effectively prohibit the use of anycast as a service distribution
mechanism.
Why was this prohibition was put into the specification?
My recollection was that it was a combination of issues.
1. ICMP and pmtu discovery not working, means you can't use it as a
source
2. We don't have a protocol that allow hosts to tell the routers which
anycast addresses they want to receive
3. TCP and higher level protocols might be confused when things change
due to routing changes
The "rebuttals" to those could be
1. Add text to say "SHOULD limit packets with an anycast source to 1280
bytes". Add note that ICMP-based tools don't work with an anycast
source address.
2. Any IGP (RIPng, OSPF, etc) can be used for this. (There was also some
work to extend MLD to do this). But this really isn't a need to
architecturally prohibit hosts from being anycast receivers; if/when
there is a protocol by which they can advertise things, they should
be able to use anycasts.
3. From the IPv6 WG we can point out that this is an issue, but leave it
to some other WG to solve (whether it is solved by a protocol
mechanisms, or operational constraints such as not using TCP with
anycast, or ensuring that routing stays stable enough)
Erik
--------------------------------------------------------------------
IETF IPv6 working group mailing list
ipv6@ietf.org
Administrative Requests: https://www1.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------