On 2007-06-27 14:17, Pekka Savola wrote:
Catching up on email..
On Thu, 21 Jun 2007, Brian E Carpenter wrote:
...
I see only downsides (unnecessary costs and useless policy discussions)
in treating this as anything but a purely technical matter. Let's leave
the policy discussions for matters where fairness and route scaling
are at stake. ULAs are plentiful (so there is no fairness issue) and
not WAN-routeable (so there is no route scaling issue).
If we don't do this, ULA-C has no noticeable advantage over PI
and we should just forget it IMHO.
Brian, it would be useful if you could elaborate a bit on "not
WAN-routeable". Some people seem to consider that as one of the major
reasons how ULA differs from PI, and I guess that leads to conclude that
ULAs have no significant impact on DFZ whereas PI might.
Specifically, I do not see any (feasible) way how router vendors could
by default drop ULA packets at some (undefined) border. This is for two
reasons; 1) it is not algorithmically possible to define which
interfaces form "ULA border" where such a filter should be applied
automatically, and 2) most vendors seem to have a policy that changes to
default behaviour (could affect existing deployments) are unacceptable
or at least very strongly frowned upon.
Why do you believe ULA addresses are intrinsically not WAN routable? Is
there something I'm missing?
We can argue about the meaning of "intrinsically" I guess. But what I mean
is that they are /48s and I don't expect to see /48s routed globally.
Architecturally, they are certainly routeable (and so are /128s).
But I am sure they will be filtered.
Brian
--------------------------------------------------------------------
IETF IPv6 working group mailing list
ipv6@ietf.org
Administrative Requests: https://www1.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------
