[AusNOG] IPv6

[Robert Hudson]

On 1 April 2015 at 00:49, Chris Hurley <chris at minopher.net.au> wrote:


> While I appreciate the horse has well and truly bolted the BIGGEST mistake
> in IPv6 was not making it backwardly compatible with IPv4.


(Let's for a moment forget that IPv6 solves more than just the problem of
the size of the IPv4 addressing scheme).

IPv6 being backward compatible with IPv4 isn't actually the issue (just
about all devices which speak IPv6 can also speak IPv4)  - IPv4 devices
being forward-compatible with IPv6 is actually the problem.

Decimal and hexadecimal notation for IP addresses is just shorthand for the
binary behind them.  IPv4 addresses can actually be expressed in
hexadecimal notation (as can IPv6 addresses be expressed in decimal
notation - dumb as that might be)  In the end, you're looking at binary
mathematics in the back end, a 0 or a 1 per bit of the IP address space,
and the only difference between an IPv4 address and an IPv6 address is the
length of the string of bits.  We only use decimal notation and hexadecimal
notation for ease of human reading - computers themselves, which actually
have to operate the address space, don't care that "IPv4 addresses are all
numbers, and IPv6 addresses have numbers and letters".

So how do you get a piece of kit designed only for 32-bit IP addresses to
allow more bits into the 32-bit IP address field, and thus make it
"backward compatible"?

Not feasible - and that was already determined by minds greater than mine
and yours.  Do you really think that if network engineering minds who came
up with IPv6 couldn't have made the transition smoother without sacrificing
capability, that they wouldn't have done so?


> If your a
> purest IPv6 is great but 'they' forget about legacy devices and the IDtenT
> issue. You might be Steven Hawkins but you have to bring the people with
> you, or at least allow for their level of understanding and willingness to
> pay.
>

Nobody forgot about the legacy issues.  Nobody is advocating that we shut
off IPv4 today, or any time in the future - but rather that we get IPv6 up
and running alongside IPv4, so when they day DOES come to shut off IPv4,
we're not suddenly rushing to test and implement IPv6 with a rapidly
approaching deadline

ID-TEN-T issues?  IPv6 auto-configuration is there exactly to address that
at the user-end.  At the network-design/configuration/operations end?
That's why we use hexadecimal notation rather than binary (easier for the
wet computing devices to comprehend), and that's why we try to subnet on
nibble boundaries, and it's also why we get to use the :: notation to
shorten strings of zeros.


> Which to be blunt is zero $.
>

Show me where you can buy modern/current IPv4-only equipment cheaper than
IPv6-capable equipment with the same features?  $50 consumer CPE can now
route IPv6, and if anything, that price of such devices has only come down
over the same time that IPv6 has been seriously discussed and offered as a
feature on this kit.

Move up into bigger networks/bigger kit, and whilst I don't run an
ISP/carrier (I do help run a global network for a corporate enterprise) and
I don't recall ever paying a single red cent extra for kit that was IPv6
capable (or even for IPv6 services where our carriers have been able to
offer them) over IPv4-only gear.

Second mistake was not making the NBN IPv6 only. This was a chance to draw
> a line in the sand, rather than say for an extreme example "You can run
> AppleTalk over it". Which under the current spec you technically can, mind
> you Apple has dropped this ages ago from their support protocols


The NBN was designed as a layer 2 network.  Layer 3 and above (except for
services such as the satellite connections, pretty much an edge-case, and
only through requirement, not desire) are out of scope for the NBN - if you
invent ChrisHurleyTalk and have enough nodes that want to use it, there's
nothing to stop an RSP from allowing it.  Are NBNCo offering satellite
services with IPv6 addresses?  I hope so.


> So given we have run out of IPv4 addresses what is the NBN proposing?
>

Nothing - it's not their problem (MAC address exhaustion could be,
different story, already, as I understand it, largely resolved).  Except in
edge cases, NBNCo will have no control over (or interest in) the addressing
used by RSPs.


> Or is the new NAT scheme the saviour? Rather than making some real hard
> calls.
>

Right now, it seems that some ISPs/carriers are deciding to invest in
NAT/PAT (or CGN, if that's really a thing) rather than IPv6.  Why?  I'd
have to suggest cheaper than a full IPv6 implementation (and given that
they'd have to be giving IPv4 IP addresses alongside IPv6 addresses
anyway), it's been selected by many as the path of least resistance.  ISPs
who do offer IPv6 (the few of them there are at present) still have to
offer IPv4 addresses to their customers anyway - I believe we've all agreed
that at this stage, an IPv6-only Internet connection won't work as
customers would require.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.ausnog.net/pipermail/ausnog/attachments/20150401/ac9a4f94/attachment.html>