[ppml] IPv6, thread diverging from 'policy' to 'background info'...

[Dave Diller]

> The problems I see are:
> Currently I am getting service from 2 ISPs (multi-homed), and 
> for me to use V6, both of them will have to be
> properly setup.


This seems to be the chicken-and-egg thing again... Maybe I'm missing
something, but I'm not clear on why everyone thinks "I can't *possibly* do it
till my provider does it".  With that kind of viewpoint, nothing'll ever get
done as everyone is waiting for everyone "one step up" to do it, and the larger
providers are naturally going to be moving much more slowly on deployment than
some of the smaller ones.... larger, more complex network means more layers of
formality to major network changes, as well as more things to test, yadda yadda
yadda.

My current status is that I received my /32 last month, and I have a copy of
O'Reilly's "IPV6 Essentials" on my desk that I've opened a couple of times - so
I'll freely admit that I'm on the steep end of the learning curve here, looking
to test and deploy this year.  So I'll ask for input/correction of the below
from those "grizzled veterenas" of v6, as I ain't one yet:

As I understand it [see OReilly quotes below], an ISP can receive a /32 from
ARIN, and then said ISP can use 6to4 to get native v6 working in their own
network, and encapsulate traffic onto the existing v4 network to get to other
v6 sites - which means their upstream IS NOT REQUIRED to run IPv6.  Does this
not work in practice?  

I'm certain it would be CLEANER if your upstream could give you a native V6
pipe, but this would at least get YOU empowered and running on YOUR network in
the interim.  What am I missing?  What is the source of the seemingly common
perception of "I must wait for my upstream to do it first!"?

Pulling out said O'Reilly (are there better resources besides the RFC's
themselves?  I waded through several of those around Thanksgiving and I'm sure
another go-round would help)

-----

Page 229, Tunnelling Techniques:

Tunnelling mechanisms can be used to deploy an IPv6 fowarding infrastructure
while the overall IPv4 infrastructure is still the basis.  Tunnelling can be
used to carry IPv6 packets by encapsulating it in IPv4 packets and tunnelling
it over the IPv4 infrastructure.  For instance, if your provider still has an
IPv4-only infrastructure, tunnelling allows youto have a corporate IPv6 network
and tunnel through your ISP's IPv4 network to reach other IPv6 hosts or
networks.

It then refers the reader to RFCs 2473 (Encapsulation with IPv6), 2893
(Automatic and Configured Tunnelling), and 3056 (6to4).

-----

Now, ignoring issues of getting the right versions of code and/or hardware to
HANDLE this, it seems this is the Best Current Practice to achieve v6 when your
upstream isn't able/willing to hand it to you natively.  Is this a correct
statement?

> Maybe there should be a policy where as if you have a V4 class C, 
> then a V6 equivalent would be free if you moved.

If you have an ISP doing IPv6, and you want to, they'll give you a /48, even if
you only have one box that wants to speak v6.  No policy required.  That's the
way it's designed.



> 
> The real problem here is not that people / smaller companies don't want to go to V6, it's that some big providers feel they
> won't have a product (IP addresses) that are scarce enough to make people buy them anymore at the rates they are paying now.

I have no relevant comment or experience with that opinion as to the "show
stoppers" of V6.  I don't charge for IPs.  At all.  But you sure better be able
to justify 'em, and you're cut off when you can qualify for an ARIN block. 
"Free milk" doesn't mean you get the whole darn cow forever.



-dd