[colug-432] IPv6: 20 years old, 10% deployed

[Rick Troth]

On 01/04/2016 11:22 AM, Rob Funk wrote:
>> > Speaking of tunnels, the resistance to them is surprising.   ... 
> "Try out the future of the internet!" doesn't cut it, huh?  :-)

Egggsactly.


> Eventually there will be more and more services that will be
> accessible only by ipv6, but for a long time ipv4 will be considered
> necessary for any service to be considered "real".

And for my part, I have tried to create a few. Mixed results.


> As I recall, one of the ipv6 tunnel services has a page listing some
> things that are accessible only by ipv6, but they're all pretty
> gimmicky and none of them have any technical reason not to exist on
> ipv4. So I can see why such a list wouldn't be persuasive.

Perhaps this one?

        https://www.sixxs.net/misc/coolstuff/


> And tunnels specifically are a pain. Why would anyone put up with 
> that pain if they don't need it and aren't interested in the technical
> experimentation?    ... 

Sure, but what I don't get is how an IPv6 tunnel is any more of a pain
than VPN.
I've found the latter to be way more trouble. (Compared to 'aiccu' for
SixXS. I completely agree that HE tunnels are a pain because ya gotta
re-do them manually every time you land on a new IPv4 lease.)

I use IPv6 in lieu of VPN where possible. (Sadly not always "possible".)


> I've also run into problems with important servers having broken ipv6
> connectivity. In particular a bunch of Ubuntu update servers have been
> working only over ipv4, even though they have ipv6 addresses. I ended
> up knocking down the precedence of those ipv6 addresses in
> /etc/gai.conf to fix the problem.

Ouch!

I find myself removing some IPv6 addrs from DNS because even if they
work (from the target end) some weak link in the chain might keep
end-to-end connectivity broken.

And there was that resolver change (five years ago or more) where AAAA
records were requested before A records. Violates the principle of least
astonishment. I don't remember how it got fixed; it simply went off my
radar.


>> > Speaking not of tunnels, when I got native IPv6 at home (having used the
>> > tunnel before), things were weird. ISP "camps" me on the lone IPv4
>> > lease, but either they or my router insist on changing the IPv6
>> > sub-leases with much higher frequency. The effect is that IPv6 native at
>> > my place is client only, useless for servers.
> Any chance it's just the kernel doing the ipv6 address privacy thing
> that I ran into earlier in this thread?

I suspect that's it (or a similar trick done in a different layer) but
haven't found the knob to turn to switch that off.

-- R; <><