How to force rapid ipv6 adoption

Owen DeLong owen at delong.com
Sat Oct 3 20:08:25 UTC 2015


> On Oct 2, 2015, at 13:45 , Todd Underwood <toddunder at gmail.com> wrote:
> 
> On Fri, Oct 2, 2015 at 2:07 PM, Owen DeLong <owen at delong.com> wrote:
>> 
>> None of them does what you propose — Smooth seamless communication between
>> an IPv4-only host and an IPv6-only host.
> 
> i view this point/question as an assertion by owen as follows:
> 
> "it was never possible to design a smooth transition and that's why we
> gave up on it."
> 
> furthermore, it's a also the following assertion:
> 
> "it was never possible to expand our address space while allowing for
> an actual migration."
> 
> if you believe that, then you end up in advocacy land.  if you don't
> believe that but you see lots of people who gave up on the design
> process early, then you understand why we're here.
> 
> v6 was designed without a migration plan and it wasn't believed to be
> important, or possibly wasn't believed to be possible.  but there was
> never any pressure to use v6 because v4 worked well and we had plenty
> of addresses.  we still have plenty of addresses and although they're
> no longer ~free from quasi-governmental organizations they're way
> cheaper than the cost to implement v6.  so we're still going to use v4
> ~forever.

OK, so if you think those are assertions rather than fact, it should be pretty
easy for you to disprove them by presenting an example of a workable
solution.

> 
>> 
>> So, please, Todd, explicate exactly how you would achieve that stated
>> objective… What could you do differently on the IPv6-only host side that
>> would allow smooth seamless connectivity to/from the IPv4 host while still
>> providing a larger address space?
> 
> it sounds like you're interested in having the engineering
> conversation that should have been had ~15 years ago.  me, too  15
> years ago.  sigh.

I’m willing to have it now if you’re up for it. So far, all I see is handwaving
claiming that it wasn’t had 15 years ago or that the fact that the conversation
15 years ago resulted in a decision that it simply wasn’t possible was somehow
incomplete rather than a recognition of the facts at hand.

I’ve given quite a bit of thought to it actually and I admit I haven’t been able
to come up with anything better than what we have in terms of migration
strategies.

> i know owen is now just trolling because he's threatened by the idea
> that there might be something wrong with ipv6, but the reality is that
> none of this was necessary.  ipv6 might have been done differently
> with a different header format and different choices around migration.
> routing could have been done differently to try to preserve end-to-end
> but still splitting locators and identifiers (which i know that dave
> meyer thinks might not be possible but i'm still more sanguine about).
> we could have explicitly made smooth migration an engineering
> requirement just as much as "more addresses" were.

First, this is absurd. I’m trying to engage you in a productive discussion, despite
your best efforts to avoid one. I’m the first person to admit that there are a
number of things wrong with IPv6, but there are also a lot of things wrong
with IPv4 and any other human invention throughout history.

However, none of what you propose above solves the problem at hand…

How does an unmodified IPv4 host accept a packet from a host with only
a 128-bit address and reply to it from it’s 32-bit address using a packet
format (IPv4) that only supports 32-bit addresses?

I agree that it would have been ideal (for other reasons, actually) if the
IPv6 packet had a 32-bit field for “Destination ASN” in it so that we
could have populated that field at the first DFZ router and then
let the packet get routed through the IDR area using just the ASN
tag. Unfortunately, that didn’t happen. (Of course now we have lots
of networks that, for reasons passing understanding, have deployed
ASNs in an incompatible way where they have multiple separate
collections of prefixes with distinct routing protocols using a single ASN).

Again, if you have a solution… An actual solution, present your proposed
packet header. Tell us how it would work. Tell us how the IPv4-only host
with no software modifications would be able to accept connections from
and respond to a host which has no unique 32-bit identifier available to it.

> we didn't.  that's fine.  so we got a disconnected network that some
> things can talk to and others can't.  and we put the full burden all
> the way to every edge.  and now we have conversations about how to
> upgrade home cpe everywhere.  it's tedious and boring and dumb but
> it's the direct result of every decision we made and how we
> prioritized things.

Yet you still haven’t presented an actual workable alternative. Lots of people
smarter than me have also pondered this question and failed to come up
with a workable alternative.

It’s not that nobody wanted what you describe… It’s that we couldn’t find
a way to implement it. If you have a solution, please present it.

If you don’t, then please stop insulting everyone as if you somehow know
better just because you can.

> so, for clarity, this "how do you magically enable smooth migration
> now that we didn't prioritize it in the protocol design" question is a
> bogus red herring.  the answer is:  "you prioritize it in the protocol
> design".  i assume smart people can see that.

It was originally prioritized in the protocol design until people much smarter
than I am with much more experience and a great deal of math decided that
it was mathematically impossible to do so.

Again, if you have a real answer to the question, please provide it. Perhaps
we can design something like what you propose and make it work. Perhaps
not, but at least we’ll know where we failed.

If you don’t have a real answer, then please recognize that many people did
try to come up with one and couldn’t. That there’s actual math that says it’s
not actually possible (I’m the first to admit the math could be wrong, but it’s
not my math and I’m not that skilled in the math involved to say one way or
another.).

If you’re so much better than everyone else who looked at this problem that
you found a solution where nobody else did, then, again, I implore you to
share it with the rest of us.

Otherwise, how about recognizing that a large number of people did the
best they could with what they had at the time and we now have a protocol
that works well enough, provides a sufficiently large address space for
many more years of internet growth, and will eventually be the internet
protocol.

Once upon a time, IPv4 was this incompatible protocol that could only
be spoken by some of the hosts on the NCP internet, you know.

> owen:  i understand you like v6 and that it's important to you.  that
> doesn't mean it's perfect and it doesn't mean we couldn't have done
> better. stop being so hostile and so threatened and try to listen a
> bit.  or don't.  whatever works for you.

Actually, I like the internet and being able to continue to deploy it to
far reaches of society is what is important to me. I recognize that IPv4
isn’t going to cut it and nobody, including you, has shown me a viable
alternative other than IPv6.

So… I’m not being hostile. I’m not threatened at all, and I’ve been
listening. The problem is that you are talking a lot without saying anything.

Todd: I understand you probably don’t have a real solution and that
you don’t want to publicly admit that because it might be embarrassing
after all your handwaving. It’s OK, I don’t blame you for that. However,
I have listened to everything you said, including a number of erroneous
assumptions about my position and where I am coming from which
I find mostly amusing, but mildly harmful. I certainly understand how
that perspective has corrupted your thinking about my words and I hope
now that I have offered you a more accurate perspective, you will take
the opportunity to review what I have said in a more accurate light.
If you have a real solution, please share. If you don’t, that’s OK, but
stop insisting that everyone else share your belief in magic and
recognize that we need something with more than 32-bits of identifier
space and we need it yesterday. We’re already way behind, so if
you have a superior solution, let’s get it out there and see if we can
get it moving. If not, how about you give the rest of us a hand 
implementing the solution we do have, ugly as it may be, instead
of just snarking from the sidelines about how it’s never going to work?

Owen

> 
> cheers!
> 
> t
> 
>> 
>> In any case I'm giving up on that conversation. And this whole one. It goes
>> nowhere.
>> 
>> And this is why v6 is where it is: true believers. Instead of a simple,
>> practical matter of engineering a transition we got 15 years of advocacy.
>> 
>> If it’s so simple, why do you continue to refuse to explain the process?
>> 
>> Owen
>> 
>> 




More information about the NANOG mailing list