[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [GLOBAL-V6] IPV4 to IPv6 migration
Brian,
On Jun 2, 2008, at 2:33 PM, Brian E Carpenter wrote:
On 6/1/08 9:36 PM, "Brian E Carpenter"
<brian.e.carpenter@gmail.com> wrote:
You *NEVER EVER EVER EVER* NAT in IPv6.
to say it quietly, there's never any reason to use nat in v6.
the only valid reason in v4 is shortage of addresses,
No. In addition to providing additional address space through the
use of
non-uniqueness, people also use NAT to provide for provider
independence.
Since IPv6 uses the same routing technology as IPv4 and it is
essentially
just as hard to renumber a site in IPv6 as it is for IPv4, it
follows that
this use of NAT will remain.
No it doesn't, which is why the current discussions in the RRG matter.
I'm not clear what you're saying "no" about. As you are fully aware,
currently IPv6 does indeed use the same routing technology as IPv4,
thus you have a choice:
a) get provider independent space and find somebody to route it
b) get provider aggregatable space and renumber if you change providers
c) use ULA and NAT into provider aggregatable space
Choose wisely.
Discussions in RRG _may_ matter, but as far as I can tell, they've
made little progress particularly in the area of reaching any sort of
consensus on pretty much anything.
And please don't encourage IT departments to believe that adding and
removing IPv6 prefixes is as hard as renumbering an IPv4 site.
(Note, I didn't say it's easy, just that it isn't as hard.)
I'm merely repeating what folks in some large scale enterprise IT
departments have told me regarding renumbering both IPv4 and IPv6.
However, perhaps things have gotten better now that DHCPv6 servers are
becoming available.
Summarizing what they've told me, in general you can make renumbering
easy or hard regardless of whether you are using IPv4 or IPv6. Network
management, firewalls, filters, and DNS configurations appear to be
where the 'fun' lies. IPv6 makes things a bit harder and easier as it
is much easier to mistype an address literal, but there are
(generally) less places you have to type them.
In addition, in both IPv4 and IPv6, a renumbering event implies
disrupted communications (particularly since the actual IP addresses
are 'cached' in applications) and some folks care about that.
It's about
time we started looking at IPv6 as an opportunity to develop new and
easier practices, rather than trying to force IPv6 into the IPv4
model.
Perhaps. I suspect, however, most large network operators
(particularly the commercial ones) will say "you first".
Of course, the alternative would be for everyone to use provider
independent
address space, but we're all aware of the fact that this won't scale.
Indeed. The PI model is very close to its sell-by date.
"We learn from history that we learn nothing from history." -- George
Bernard Shaw
Regards,
-drc
