Moving to IPV6

[Mike O'Dell]

Sigh...

NAT is not going away.
Believing otherwise is fantasy (even if it seems attractive).
In fact, one hears the term "Carrier-Grade NAT",
at which point the blood runs ice-cold.

Flat addressing cannot possibly scale.
We have the worked example now, and having a lot more
of a problem we cannot solve will not
make things better any time soon.

Global reachability is REQUIRED, but one needs to
contemplate what that really means given that
Flat Addressing Just Doesn't Work.

Flat addressing is not required for global reachability.
Hint: space can be "locally flat" but still be located
on a complex manifold. think relativistic gravitation vs
Newtonian gravitation.

"NAT" is the unfortunate (if accurate) name given to
"create an arbitrary abstraction boundary" in IPv4 & IPv6.
the name is based on the confusion of architecture
with the implementation given the realities of IPv4.

the first two things an "internetwork
protocol" must be able to do is (1) designate the
participants of a packet exchange and (2) provide for
the deliver of said packets between participants.
When those two are explicitly Not Considered in the
design of IPv6, it's hard to understand what *was*
considered and why it matters.It's rather like
designing a radio system without providing for tuning
or considering propagation.

the fact that IPv6 was *explicitly* created without
regard for addressing or routing is strong evidence
of the profound naivete driving its design. (I say that
as one of the afflicted.)

the implementation of NAT for IPv4 and IPv6 *is* gruesome
but it's not the fault of the concept. as for uPNP and similar
other abominations, you won't hear any defense from me.

A surprising amount of what's been piled  on top of IPv4,
especially things "broken by NAT", don't really
"work from first principles", but rather "fail to fail" often enough
to mislead many people into believing they work
(even some who should know better).

Whether this is a failure of the underlying architecture, a
misapprehension of what that underlying architecture can really do,
or just mind-bogglingly-bad protocol design is split about 20/30/50
percent, respectively (to about one digit of precision).

sorry if this sounds horribly cynical, but i've been watching the
sausage being made for over 30 years now and have helped turn the crank
on more than a few occasions. i wish i could offer a silver bullet,
i don't know of any.

     -mo



On 2/9/11 8:52 PM, Josh Smith wrote:
> On Wed, Feb 9, 2011 at 4:48 PM, Mike O'Dell<mo at 131.ccr.org>  wrote:
>> NAT is not "nasty" - it's a fundamental architectural concept
>>
>> the nastiness of doing NAT is a reflection of
>> the decrepitude of the underlying architecture
>>
>> creating arbitrary abstraction boundaries is a beautifully
>> natural thing to do, and the fact it's ugly in IPv4 and IPv6
>> is purely a reflection on IPv4 and IPv6, not the other way around
>>
>>    -mo
>>
> NAT is a nasty hack, all of the technology developed to make it
> palatable for the consumer, uPNP nad etc are equally nasty.  Wile I
> agree ipv6 has some of its own other problems.  I look forward to not
> having to worry about NAT and the crud that accompanies it.
>
> Just my 2c.
>
> Thanks,