[clug] IPv6

[Paul Wayper]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On 01/12/2011 11:47 PM, David Austin wrote:
> Perhaps DoD early adoption is a reflection of their propensity to
> adopt sh*t standards all over the place...
> 
> (not impressed that IPv6 is not backwards compatible)

And how would you make it backward compatible?

And just what do you mean by 'backward compatible'?  Do you mean 'my Linux 2.2
machine has to be able to work in an IPv6 world'?  Do you mean 'my IPv6-only
website has to be accessible to IPv4-only machines'?  Do you mean 'I shouldn't
have to learn new addressing schemes, protocols and networking methods, the
old ones should just work by magic'?  Or what?

What you're probably talking about is 'I want my IPv6-only machine to be able
to access IPv4 addresses with no additional hardware, routing or proxies'.
And the bad news is that that's really just a fantasy.  The IPv4 IP packet is
fairly prescriptive about what goes where, what it all means, and its
extensions are fairly limited.  You can't just bolt on an option that says
'oh, yeah, and the other 96 bits of the source and destination are in this
other part of the packet'.  The IPv4-only server you're talking to is going to
toss those bits away and you're never going to get your response.

This is why people run dual-stack.  You get an IPv6 address and an IPv4
address and your computer uses whichever it can to get your requested website on.

Reading things like this:

http://www.networkworld.com/news/2009/032509-ipv6-mistake.html

give one the false impression that the reason you can't have backward
compatibility is because some back-room boffins didn't think hard enough about
it.  I personally doubt this.  There are plenty of good reasons for not being
backward compatible, starting with it being an awful kludge however you do it,
continuing through trying to avoid some of the naive assumptions about how
networks theoretically worked, and ending at the practical abilities of
modern-day technology and modern routing implementations.

I'm more than happy to hear the wherefores and whys of how IPv6 was developed,
but I refuse to accept that backward compatibility was left out by pure dumb
accident, and now there are labs full of network engineers slapping their
foreheads and going "oh, of course, we forgot to specify the 'compatible with
IPv4' bit - it would have been so easy!".

(From my own experience, I also think that the slow pace of IPv6 deployment
isn't because a whole bunch of network engineers are sitting in the back rooms
saying "hmmm, Management wants us to implement IPv6, but it sounds a bit
difficult, let's play Go instead" - partly because that's now how organisation
power structures work, and partly because that's not how most network
engineers I know work.  It's more likely to be the engineers saying "when are
we going to implement IPv6" and management asking "well, can we survive
without it?", to the effect we see today.)

Have fun,

Paul
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.11 (GNU/Linux)
Comment: Using GnuPG with Fedora - http://enigmail.mozdev.org/

iEYEARECAAYFAk0uok8ACgkQu7W0U8VsXYJ4oACgqnhKJMDbbZi653eOCazjtMlU
hhMAoLLZ81kpqCAkIBXyuc5uZnwLEN97
=Mgj5
-----END PGP SIGNATURE-----