... oversight.
I don't think this is an oversight, I'm pretty sure this was
intentional. However, since in practice the BGP best path selection
algorithm boils down to looking at the AS path length and this has the
tendency to be the same length for many paths, BGP is fairly useless
for deciding the best path for even low ambition definitions of the word.
that depends on what you want to use it for. not all routes are transitted.
for f-root we do limited-horizon peering in all locations outside the SFbay
area, and bgp's path selection machinery almost never has any work to do.
i'm a little more worried about j-root which receives transit everywhere and
is somewhat at the mercy of not only its own ISP's but other ISP's as well.
however, "diversity is good". i'm glad that different roots are different.
(And some IPv6 roots wouldn't be bad either.)
there are several. see www.root-servers.org. (now if we can just
advertise.)
Just for fun, I cooked up a named.root file with only those IPv6 addresses
in it. This seems to confuse BIND such that its behavior becomes very
unpredictable.
hmmm. that configuration works fine for me here.
And only 2 of the 4 v6 addresses are reachable as one isn't
advertised at all
those issues are probably going to influence which AAAA RR's go into the
root-servers.net zone and the named.cache files.
and the other as a /48 which are heavily filtered.
not according to the RIR's. at least in ARIN's case the micro-allocation
policy seems to have met with approval by the membership, which is why we're
using a /48 for f-root. if this is a bad idea because all kinds of ISP's
won't be accepting such routes even though they seem to be grouped together
in a place where a different prefix filter could be employed, then you ought
to tell the RIR's this and get the microallocation policies altered or torn
down. (i personally don't think a /35 route with just one host in it makes
much sense, but i guess ipv6 has a lot of space to burn on stuff like this.)