Ask HN: So why does IPV6 appear to be irrelevant? Will this change?

8 points by hoodoof ↗ HN
As a user, developer or sysadmin, I appear to be able to completely ignore IPV6.

Is this likely to change any time soon? When will it start to be something that I must pay attention to?

13 comments

[ 3.4 ms ] story [ 43.0 ms ] thread
Facebook are now using IPV6 only for all their internal networking ( http://www.internetsociety.org/deploy360/blog/2014/03/facebo... )

While it's possible to get by on IPV4, most of what we're doing is workarounds. DHCP is a workaround (although a useful one), NAT is a workaround, at least some VPN use-cases are workarounds (e.g inter-office connection between private subnets)

A lot of problems go away when you have essentially unlimited address space to play with.

The downside so far is that very few providers offer native IPV6. This is frustrating but manageable. There are tunnelling providers such as Hurricane Electric ( http://tunnelbroker.net ) and SixXs ( http://sixxs.net ) Hurricane Electric are great, and they offer free self-study and certification. I was unable to open an account at SixXs.

My advice would be to start experimenting with it now. This will give you a headstart over the rest of the field as IPV6 becomes more popular. If you're working with new infrastructure, experiment with IPV6 only. If you can solve those problems, your skills will be in high demand

> My advice would be to start experimenting with it now. This will give you a headstart over the rest of the field as IPV6 becomes more popular.

I remember specifically a coworker of mine in 1996 saying that I had 4 years to learn about IPng as it would be replacing everything by the Year 2000.

I guess the work we had to do stopping planes falling out of the sky and holding back the Y2K zombie apocalypse delayed that somewhat. :)

> When will it start to be something that I must pay attention to?

As soon as people like you decide it is. Chicken and the egg. As long as things keep working the way they are, don't expect people to move too quickly.

IPv6 is (getting more) relevant in parts of the world, because (duh) we're running out of IPv4 addresses. All the backbones and most data centers are already fully IPv6 capable, don't know about cloud providers though. If I rent a bare metal or VPS i can expect full IPv6 connectivity, at least with all the big vendors here in Germany.

We have ISPs here (Unitymedia, one of the largest German cable companies) that have run out of IPv4 and started handing out IPv6-only contrancts to all new customers. Those customers have to access IPv4 resources using DS-Lite /CGN (carrier grade NAT), which is a horrible workaround with a lot of issues. Funny enugh, the same ISPs won't give IPv6 to existing customers because every change will probably lead to increased support costs and make the customer unprofitable.

There was a great talk by a Cisco guy at last year's FOSDEM explaining what to expect with IPv6 and why CGN really, really sucks. Video: https://archive.fosdem.org/2014/schedule/event/no_more_ipv4/

So yes, everyone who serves stuff on the internet should care because you already have or will soon have customers that only have native IPv6 and are best served reliably over IPv6.

I work for a very large company that is supposedly "leading the charge" in the IPv6 space.

Three years ago all groups were supposed to have converted completely to IPv6 by the end of the year. Never happened.

Back then the exhaustion of IPv4 addresses was "any day now". I guess someone must have found a closet with a bunch more of them somewhere.

I think for us operations types IPv6 is still too complicated. I mean when someone says "hey, what's the IP of that web server?", saying "192.168.13.129" is easy. You can even memorize an address like that. "fe80::2bdd:d4c5:f093:300a", not so much.

Besides the IPv4 address exhaustion problem I don't think anyone has made a compelling argument for IPv6 yet. At least not on the ground.

>I guess someone must have found a closet with a bunch more of them somewhere.

I lived in such a closet one summer a couple of years ago. If I'm getting this terminology right, they had an entire 16-bit block to themselves (65k addresses).

Company I just left after 12 years had an entire /8 and a /16. One of the projects I worked on in 2014 was moving a lot of the stuff out of the /16 (which had been owned by a subsidiary that was bought almost 20 years ago) and into the larger subnet.

I heard rumors that they were going to renumber out of the /8 and start using RFC1918 addresses for all "internal" (non-Internet-facing) stuff in the next couple of years.

> I think for us operations types IPv6 is still too complicated. I mean when someone says "hey, what's the IP of that web server?", saying "192.168.13.129" is easy.

There's a cure for that, it's called DNS. But seriously, building IPv6 networks feels strage for a few hours, after that you just wish IPv4 would disappear because it's clumsy and ambiguous in comparison.

IPv6 leaves so much room to logically arrange your (and the world's) whole network in one namespace without any RFC1918, making routing and firewall configuration really easy and elegant. Soon you will be able to know a system's VLAN/function/location just by looking at certain parts of the address.

Memorization is important. Although so many more provisioning/procuring processes are automated than they were during the Internet's rapid growth phase in the West, engineers and other professionals still have a great need to handle the addresses, and to do so without frequently needing to look them up.

I observed that it took a little while for the majority of network-touching professionals to become accustomed to memorizing an IPv4 address in one or two glances. I believe doing the same for an IPv6 address is possible. As this happens, the adoption rate will accelerate.

When will it start to be something that I must pay attention to?

As long as you're not doing anything sufficiently interesting from a network perspective (a home LAN and hosting webapps at VM providers do not count) you've got a least a decade more. No matter how far things advance, in 10 years we'll probably still be supporting stragglers with the ever increasing ball of complexity we use to avoid going v6-only.

You can ignore it because you are blissfully ignorant. This is a good thing, because it indicates that most cases the transition to IPv6 is fairly seamless and unnoticeable.

Meanwhile on the network engineering side of things, most of our traffic is IPv6[1]. I deal with it daily.

[1] Disclaimer: This varies by provider and client base, but is true in my case.

For the most part if you're using modern software you can ignore it and it'll just happen automatically. Your software probably already supports it, so if you just turn it on it will just work.

The main user-facing and developer-facing difference is a /128 address space. The rest of the differences can mostly be ignored or learned slowly by most people who aren't hard-core network admins.