Thus spake "David R. Conrad" <David(_dot_)Conrad(_at_)nominum(_dot_)com>
Those who fail to learn from history...
Very true. How many prefixes are in the swamp?
While we've learned from the experience with the swamp, we've also
learned that non-portable addresses can't be the *only* type available.
You can't simply banish multihoming by fiat, as much as current registry
policies are geared to do so.
What'd be better is for SOME organization, perhaps IANA, setting up
one
provider-sized block of addresses for early adopters to USE.
Unless you are purposely trying to create a new (potentially *much*
larger) swamp, causing v6 to run into the _exact same_ routability
issues that exist with v4, setting aside a block for "early adopters"
would seem to be a move in exactly the wrong direction.
Giving a single prefix to each AS would not create a "swamp" in the same
sense we have today. A world with 65,536 IPv6 routes would hardly be
unmanageable -- it's far better than where we're at today with IPv4.
As Brian said, get address space from your upstream provider. If your
provider doesn't support v6, find another. If you can't find another
then get used to and deal with the fact that you will have to
renumber.
Provided you're single-homed, that's a great plan. IPv6 renumbering is
a reasonable task, provided it's not done too frequently.
Here's where the general wisdom that we should all shift to IPv6
meets
the reality that SOMEONE has to ante up and provide a way for folks
to
start really working with the protocol, with REAL and routable
addresses.
Routability is defined by service providers, not TLA allocation
registries.
Indeed. The "TLA allocation registries" then should have no prohibition
on allocating prefixes which the service providers may or may not route
in the future. Providing a warning message is probably suitable,
however.
The problem here is that the TLA/SLA system is unrealistic when end
sites become multihomed. Without portable addresses, multihoming simply
does not work in a reasonable fashion.
If there is no means for multihomed end sites to get portable addresses,
that makes IPv6 inferior to IPv4 for practical use. It will be
extremely difficult to convince people to "upgrade" to a new protocol
which supports less functionality than the one they're using today.
The allocation registries merely define uniqueness,
something that does not matter if you are not connecting to the
Internet.
Uniqueness is mandatory for any two organizations connecting together,
even if they don't connect to the Internet itself. Trying to allocate
private addresses between N parties partially meshed to each other is a
simple task when N is small, but the nature of the beast is that N grows
very quickly. Without a registry allocating unique addresses, we'll end
up with NAT all over again.
Rgds,
-drc
S
| | Stephen Sprunk, K5SSS, CCIE #3723
:|: :|: Network Design Consultant, GSE
:|||: :|||: 14875 Landmark Blvd #400; Dallas, TX
.:|||||||:..:|||||||:. Email: ssprunk(_at_)cisco(_dot_)com