Re: IPv6 Pain Experiment

[Doug Barton <dougb () dougbarton us>]

Another misconception. Humans (by and large) count in decimal, base 10. 
IPv4 is not that. It only LOOKS like that. In fact, the similarity to 
familiar decimal numbers is one of the reasons that people who are new 
to networking stumble early on, find CIDR challenging, etc.

I do understand that the hex thing presents a (small) learning curve. 
But work with it for a little while and it will become familiar, just 
like IPv4 did.

In fact, once you get past a few basic concepts, the network'y bits 
should be familiar to you (pun intended). CIDR works the same way, the 
only real differences there are that a /64 is your basic network unit, 
and is roughly equivalent to an IPv4 /22 in the sense that ~1,000 hosts 
per network/VLAN is pretty much your limit. The other thing to keep in 
mind is that due to the massive size of the address space, it's rarely 
useful to allocate on anything other than a nibble boundary (that is, 
divisible by 4). There are two reasons, sparse allocation, and the fact 
that reverse DNS is much easier if you keep things in that framework.

Now I do admit that the whole RA/SLAAC vs. DHCPv6 thing is more 
complicated than it should be. Some of us fought very hard for the 
concept that SLAAC should be optional, and restricted to network and 
gateway; but we lost to the "SLAAC must be the new DHCP!" crowd. Sucks 
that you have to do both, but since you're already doing DHCP for 
end-user hosts anyway, and you're already configuring the router for the 
IPv6 network info, the marginal cost isn't really that high.

Take a look at 
https://dougbarton.us/IPv4_and_IPv6_Network_Structure_Planning-20190519.xls 
if you're interested in learning more. I have some cheat sheets that 
will help you understand assignment strategy, sparse allocation, nibble 
boundaries, etc. It also has handy calculators that allow you to plan 
for IPv4 and IPv6 networking based on the number of different 
types/sizes of offices, data centers, etc. in each region.

Enjoy,

Doug


On 10/2/19 5:54 PM, Matt Hoppes wrote:
> I disagree on that. Ipv4 is very human readable. It is numbers.
>
> Ipv6 is not human numbers. It’s hex, which is not how we normally county.
>
> It is all water under the bridge now, but I really feel like ipv6 could have been made more human friendly and ipv4 
> interoperable.
>
> > On Oct 2, 2019, at 8:49 PM, Doug Barton <dougb () dougbarton us> wrote:
> >
> > > On 10/2/19 3:03 PM, Naslund, Steve wrote:
> > > The next largest hurdle is trying to explain to your server guys that you are going to go with all dynamically assigned 
> > > addressing now
> >
> > Completely false, but a very common misconception. There is nothing about IPv6 that prevents you from assigning static 
> > addresses.
> >
> > > and explaining to your system admin that can’t get a net mask in v4 figured out, how to configure their systems for 
> > > IPv6.
> >
> > If they only need an outbound connection, they probably don't need any configuration. The instructions for assigning a 
> > static address for inbound connections vary by OS, but I've seen a lot of them, and none of them are more than 10 lines long.
> >
> > Regarding the previous comments about all the drama of adding DNS records, etc.; that is what IPAM systems are for. If you're 
> > small enough that you don't need an IPAM for IPv4, you almost certainly don't for IPv6.
> >
> > IPv6 is different, but it's not any more difficult to learn than IPv4. (You weren't born understanding IPv4 either.)
> >
> > Doug