Ask HN: WTF am I supposed to do about IPv4 (address exhaustion)? How can I help?

In order:

- Pretty much nothing. Actually I believe that people who request more IPs now, help to push the migration sooner since there's no other solution. ISPs are going to support it as soon as it becomes required.

- Yes, you need an IPv6-compatible ISP. In recent linux distros everything should "just work" as soon as the upstream ISP assigns you an IPv6 address. Not sure about other systems.

- If it supports IPv6, it will probably do that by default, so you don't need to enable anything. If it doesn't, there will be a lot of manuals soon...

- It's something users should not care about and admins should know everything about... so I doubt such document can exist. (actually no - people might want to know at some point why they can't access www.somenewsite.com, but Bob can - vendors will happily explain that to them, while selling a new copy of the system / router / ip-phone / ...)

- Request more IPs, help us migrate faster ;)

This is an admirable question, although the answer is that there's really nothing you can do. ISPs and many content providers are planning to enable IPv6 at the last minute (there is some debate about when that actually is, but that's not relevant here) and no sooner.

You can use IPv6 today by using either 6to4 on your router or Teredo on your hosts. This is probably a good idea if you're a developer so you can test your software over IPv6.

"The last minute" is whenever they ask IANA for some more IP addresses, and IANA says "No."

Seriously, though, there's starting to be some real movement on this. It's all very preliminary, and about where I hoped it would be in 2000, but there is movement.

I will second comments of this being an admirable question. Unfortunately, conserving v4 addresses probably won't help given the rate we're burning through them (10 /8s this year already; compare with a total of 8 /8s in total in 2009 [1]). Transitioning is probably the best thing we can do now. Note that the v6 transition process (which has been theoretically going for about 12 years now) should be a case study motivating the design of backwards-compatible protocols in large, distributed, and decentralized computer systems, in case we ever have to do something like this again.

While I don't run/work at an ISP, the biggest claim that seems to be made against v6 is that there is no business case for implementing it -- it's simply providing (at best) the same connectivity with added capital and operations expenses. While I think this is awfully short-sighted, it may make sense to management of some ISPs. Regardless, how can you create incentives for the entities that you depend on for internet access?

As a consumer you can start by asking your residential access ISP about v6. If they're on the ball, like Comcast, you should participate in their trials [2]. If not, ask them what their strategy is and/or when you can expect to have some form of IPv6 connectivity. If you want to be an early adopter, turn on v6 on your own equipment and use Teredo or a tunnel from Hurricane Electric or another tunnel provider. Be prepared for things to be slightly broken, and to need to navigate to special urls like ipv6.google.com in your browser/etc. But you should keep doing this even if things are broken, to help sites you care about work through the transition. If you need to buy new home networking equipment, consider looking at v6 support (though you may want to hold off until residential providers announce their plans regarding their transition technology of choice (ds-lite, 6rd, etc.))

As someone with purchasing authority at a smaller business/etc., you might consider doing much of the same as noted above, though you may have the additional option of switching providers if you feel so strongly about it. You will probably also want to think about experimenting with and eventually deploying v6 internally so you're ready to go when the v6 Internet comes knocking.

Finally, if you're a software developer like may readers of HN are, you should make your application IP-version agnostic by replacing calls to v4-era functions like gethostbyname() with getaddrinfo(), etc. and not store IP addresses as ints [3].

[1] http://www.iana.org/assignments/ipv4-address-space/ipv4-addr... [2] http://comcast6.net/ [3] http://people.redhat.com/drepper/userapi-ipv6.html

- if you're a software developer, make sure your code can handle ipv6 address strings and fields. Very few do.

- if you work at an ISP, start evangelizing ipv6 to coworkers and managers. Major ISPs have had their heads in the sand and are really counting on implementing carrier-level NAT, which brings me to...

- if you're an ISP customer, make it quite clear to to your ISP that you won't accept a connection behind an ISP NAT, and you'll switch to a different ISP if they impose one on you. Also make it clear that you'll advise anyone you do tech support for (families, friends) to do the same.

- if you program network video games or voip software, get the word out (either through offical documentation or a blog) about the evil of ISP-level NAT and how this will affect your application.

Actually, some sort of ipv4 NAT will be inevitable for ISP customers. But the really atrocious ISPs won't give you a unique block of ipv6 addresses to mitigate that.

I've seen companies request blocks of addresses solely to enable remote access to their workstations, when a single address could do the job (with port mapping or a VPN).

As individuals, there's not much we can do. IP addresses are cheap, and business won't care how scarce they are until it affects the bottom line.

If they're doing it now, it might be a good decision. Next time their request for IPs is denied, they will move remote access behind nat, or to a single ip and get new IPs "for free". It's a much better situation than suddenly finding out your request is denied and you can't scale down, because you've got your current ranges tightly packed. It will affect the bottom line much more in that case (deployment delays, higher prices, etc.)

A possible solution to make a work station, mobile or embedded device accessible (over HTTP) without a public IP address is using a relay server like http://yaler.org/

The best thing you could do is to use as many IPv4 addresses as you can... just to speed up the migration process :)

I actually like this suggestion. People don't want to switch to IPv6 because there's no business case for it. Make the situation they face more drastic, they get their their business case.

Why do you care? It's not a resource that's a massive loss once it's gone like forests or oil. It's still all there, just in use.

I don't really see a big problem with a move to IPv6 either - there are four main tripping points and all four of them aren't a problem. Imagine there's no more IPv4 space to request and you are forced to use IPv6 for your new site.

1) Can your IPv6 site get out to the IPv4 internet? Yes - you can have an IPv4 address behind NAT if it comes to it.

2) Can your IPv6 site get out to the IPv6 internet? Yes.

3) Can the IPv6 internet get to your site? Yes.

4) Can the IPv4 internet get to your site? If they have a tunnel broker. That means no. But as IPv6 takes off, ISPs will do this automagically.

So shortly, the only problem with having an IPv6 only site is that some users around the world on slower to adapt ISPs wont be able to get to it ... yet.

The hard parts are device firmware not supporting IPv6, planning and rollout and management and billing systems not supporting IPv6, and that's not something you can fix (except by becoming a consultant) but it's also not something you really care about. So Joe's Cheap Websites Ltd. management page wont let you create IPv6 DNS records. Don't use them. So printers from 1998 wont let you give them IPv6 addresses - use IPv4 on that network, it will still work.

This assumes applications are ready for IPv6. They overwhelmingly aren't. Virtually all networked applications assume that IP address resolution is going to result in a scalar integer and will break if it doesn't.

Well, you could change out all of the routers you have influence over and at least have all of your internal stuff already running IPv6...

I notice that Ubuntu Lucid Lynx and Windows 7/2008 all install with IP6 ready to go...

Would there be a business in reverse-engineering and then patching important binary applications (lacking source code) to use IPv6 instead of IP4?