Steve Deering's Hourglass presentation is still a classic. Whether you
use v4 or v6. That's a nice recollection of him and 1997.
The thing is by 2000 we had to run with IPv6 transition plan because
there was zero chance of getting a v6 that was v4 wire compatible
through the IETF. That plan depended on so many free IPv4 space that the
transition had to happen early or it would become "too late" and then
we'd end up with a gatewayed portal internetwork. That would be really
bad news for a rich ecosystem of ISPs and Application services.
Of course that was all before "the Cloud", Google, Amazon, Facebook,
..... and CDNs.....
If anything the CFOs of ISPs and network service providers should be
looking not just at the cost of IP addresses including implementation
but the risk cost of not having them. This has been a hard sell because
of the very short time horizons of current business practices and
investor demands.
best Christian
On 26/05/2020 12:03, Per Bilse wrote:
I'm certainly not opposed to making technical progress. I once
arranged a full day's workshop hosted by Steve Deering to evangelize,
but I have come to realize that it was 20+ years too early. (Amusing
anecdote: it wasn't really on Deering's radar to do these things, but
he had trouble getting a hotel room for the Munich IETF meeting in
1997, and I offered him mine. He thought that was very generous, and
gladly agreed.)
It's just practical reality, and spending more time and energy on IPv6
won't get far up any significant chain of command. By way of example,
consider the BGP3->BGP4 and CIDR transition. I once shared the
questionable honour of announcing more unaggregatable prefixes than
any other AS in the world, and everybody everywhere knew where things
were heading. Still, the project didn't gain real momentum until all
the world's AGS+ routers started keeling over due to running out of
memory, and that was with all the goodwill and cooperative spirit
inherent in an Internet where NSFNET access was the holy grail of
connectivity.
Money talks, it's that simple. Until the current state of affairs
becomes less profitable (one way or another), the current state will
prevail.
I don't know why you think I'm being critical of the KAME project; I
most certainly am not, and I specifically point out that it's probably
the most widely known and used IPv6 code base. If I'm being critical,
it's of the world at large (think big and share the love :-) The
world has arguably failed to follow up on the momentum IPv6 had at the
time, but things probably couldn't have gone any other way, and
there's still some way to go.
Best,
-- Per
On Monday, 25 May 2020, 17:24:43 BST, Christian <c...@firsthand.net>
wrote:
Dear Per,
When is too late now? The original transition plans had to be
revised at IETF around 2008 largely because it was only "just before
it's too late". So there has been a reprieve of a hard withdrawl for
an extra 12 years. But can you say now when would you know if you left
it just before too late again? Or when it is actually too late!?
There are plenty of folk who have been engaging in IPv6 issues
collectively in Fora, TaskForces, Readiness activities. ISOC has its
360 Deploy team. All take a strong interest in unpicking the detail
and distribute clue as to the detail of how to do things better. UKNOF
has also been a pioneering place that has always been open to engaging
on IPv6.
Personally I don't think we can or should talk about The Internet
going IPv6 native as if there is some kind of future NCP/TCP cut off
repeat. We need to live with and build on what we have.
best
Christian
NB Ito Jun is unable of course to defend himself and WIDE Kame now. I
would just say that he provided every opportunity for you to build on
that code and rather than sniping at it twenty five years on. It would
be great to see acknowledgement of the enormous contribution that was
made openly to the benefit of the Internet communities from that team
and build from it.
On 25/05/2020 10:17, Per Bilse wrote:
When I saw this thread starting I thought to myself "No, no, no ...
this will never end well." I decided to stay out of it, but Neil
raised an important point, namely that things are more complicated
than what meets the eye. The state of IPv6 software in the field
isn't good, and much of it dates back to the early years of this
century, or even late last century. There was great enthusiasm and a
flurry of activity when IPv6 first saw the light of day, but the
endurance of IPv4 combined with a steady stream of routine, everyday
issues meant that IPv6 started taking a back seat after just a few
years. I keep telling this story about a journalist asking me "When
will the Internet change to IPv6?", to which I replied "Just before
it's too late"; it wasn't what he was expecting to hear, but it was a
reflection of daily reality in a commercial environment, and that
hasn't changed. IPv6 remained a draft standard, accompanied by
various additional RFCs and related documents, until it was finally
consolidated in RFC8200 a few years ago; the process took nearly 20
years, and the promotion to full standard was partly prompted by an
administrative change in the RFC process.
Being stuck in the draft standard state isn't in itself unusual, and
interoperability demands from the real world usually iron out any
implementation bumps, but in this respect IPv6 has been in a
backwater. Probably the most widely known and used IPv6 code base came
out of the WIDE/KAME project, but much of it is literally 15-20+ years
old, and hasn't been updated for over a decade. Hence, by today there
are dozens, or probably actually hundreds, of slightly diverging IPv6
implementations in various states of disrepair. They trace their
origins back to RFC2460 from 1998, but each will have incorporated its
own blend of tweaks, fixes, and updates, some from RFCs and errata,
some from I-Ds, and some from in-house developments. Few, if any, are
ready for prime time in the way IPv4 is, and while they all generally
work, IPv6 simply doesn't have the tightly knit development and
interoperability history that IPv4 has.
So ... when will the Internet go IPv6 native? Maybe never. It's a
commercial issue, and there are no overarching considerations at play:
IPv4 doesn't cause global warming, and the ozone layer is safe too.
Hence, nobody is likely to spend time and energy on IPv6 until doing
so becomes a source of revenue, and that will in the first instance
mean the whole world catching up with two decades of neglect.
-- Per
