Hacker News Reader

Comment by Kranar | Hacker News Reader

IPv6 allows for so many IP addresses that you could give yourself a new IP address every minute and never worry about running out.

View on HN

You could, but does your device actually do so? Are you confident that 100% of the devices you own use privacy addresses and do not leak non-privacy addresses?

The nice thing about IPv4 NAT is that you plug a single gadget in to your ISP's connection (cable modem, ONT, whatever), you connect your devices to that gadget, and it works out of the box and has all the security properties you'd expect, even if you're a person who hasn't ever thought about security properties and doesn't know what IP is and is still running Windows XP because your word processor still works fine. It might be unclean, but the benefits of this model are immense.

Most operating systems default to using Privacy Extensions which will cycle your address at frequent intervals, unless you specifically disable it.

One problem with IPv4 is that each interface can only have one IP. With IPv6 an interface can have many! For example both a private and a public IP address!

?? multiple IPv4 addresses work fine... ip addr add 192.0.2.2 dev eth0

It's true that IPv6 requires IP stacks to support this, but nothing in IPv4 prevents IP stacks from supporting it.

> Are you confident that 100% of the devices you own use privacy addresses and do not leak non-privacy addresses?

I've heard stories, but I've never seen a modern operating system without privacy extensions in the wild. Maybe XP, but did XP have IPv6 enabled by default? As far as I know it didn't.

Every time I see a conversation about IPv6, this argument comes up. Is "it's got a gorillion IP addresses" the only redeemable quality IPv6 has?

It's less a redeeming quality and more the entire reason why we bothered to switch to a new system. So yes, that is the point.

Then, given that CG-NAT solutions have proven to work well and they don't bother most end users, it's no wonder the transition to IPv6 has stopped.

> it's no wonder the transition to IPv6 has stopped

Where are you getting that from? Looking at public IPv6 statistics from big companies shows that it's going up. Google's shows a 4.5% increase in IPv6 of total traffic during the past year[0]. Facebook's shows a 6% increase and at 55% for the US [1]. Meanwhile Mexico went from 5% IPv6 to 25% this year [2].

[0]: https://www.google.com/intl/en/ipv6/statistics.html [1]: https://www.facebook.com/ipv6/?tab=ipv6 [2]: https://www.vyncke.org/ipv6status/project.php?metric=p&timef...

These statistics are meaningless propaganda. What you should be interested in is the velocity of IPv6 adoption growth among ISPs and countries where adoption stopped all together, to know what to expect from actual adoption in coming years. (it's basically a failure btw)

> What you should be interested in is the velocity of IPv6 adoption growth among ISPs and countries where adoption stopped all together, to know what to expect from actual adoption in coming years. (it's basically a failure btw)

I don't understand. Are you telling me to look at how IPv6 adoption is among countries, where IPv6 usage growth has stopped? Are there any such countries? Furthermore, I'm not sure what you mean by "actual adoption". Every statistic I look at, as referenced in my previous comment, I see IPv6 usage continuing to grow. And you're calling that "propaganda"?

For ISPs (and mobile carriers) that are actively involved in their infrastructure, CG-NAT is more expensive to run than native ipv6 (aquiring IPv4 public addresses; tracking connection states; if all customers are on private IPv4, maintaining multiple distinct private networks with the same address space is a hassle too), so those ISPs are pushing to make services available over IPv6 (I've heard from T-Mobile USA and Jio India), but that doesn't mean they won't also run CG-NAT for customers on older devices and for services that aren't IPv6 -- it just reduces the deployment. For smaller ISPs, or where everything is contracted out, the cost savings aren't as apparent.

I don't think its helpful to discredit all the other technical advantages of v6. Address assignment and configuration is significantly better under SLAAC than DHCP, multicasting actually makes sense, and routing logic and packet format are simplified.

If you were to try implementing a router for ipv4 and v6 the v6 one would be dramatically simplier with a lot fewer convoluted edge cases.

I actually specifically do regard (most of) v6's additional features as a bug; I think most of them could and therefore should have been implemented as separate protocols (consider ipsec, which actually did get pulled out and back ported to 4). I think the protocol we ended up with suffers from horrible second system effect and its scope creep it's part of the reason why it's taken decades to get mass adoption.

Reply on news.ycombinator.com