Live data from Hacker News

16 Years with IPv6

teamarin.net

41–50 of 65 posts

Re: 16 Years with IPv6

#42
IPv6 is not a requirement in all situations. At most enterprises need IPv6 at their edge to avoid traffic flowing through carrier grade nats. Internally private IPv4, plus nats for egress are generally fine for most use cases.

Re: 16 Years with IPv6

#43
post #21
post #8

In Seattle and surrounding areas, Comcast is more or less the only game in town (with Centurylink a distant, distant second) if you're not in one of a dozen high-rise apartment buildings. For all their marketing around 'extensive IPv6 rollout', I've found their deployment to be hideously broken and their support staff deeply ignorant. Been here 3 years, moved once, and it's just terrible. My latest foray into trying…

In Seattle there is also CondoInternet - a fiber company in many of the newer development. But, I've seen an entire building on a NAT private network. It's included in your HOA.

Condointernet got acquired by Wave (branded Wave G); then they fired most of the condointernet staff it seems. Wave recently got acquired themselves so we'll see how much worse it gets.

Re: 16 Years with IPv6

#44

IPv6 is not a requirement in all situations. At most enterprises need IPv6 at their edge to avoid traffic flowing through carrier grade nats. Internally private IPv4, plus nats for egress are generally fine for most use cases.

Internal IPv6 is much cleaner though, and less prone to failures (e.g. anyone VPNing in from a coffee shop that uses the same private IPv4 range will have a bad time). If you've already gone to the effort of getting IPv6 to the edge, surely at that point it's worth pushing it all the way through.

Re: 16 Years with IPv6

#45
I'm with a regional ISP in the UK (aquiss.net) and they assign you a native /56 if you ask nicely. What surprised me was how complicated it still is to get this set up compared to the DHCP and NAT setup we're all familiar with. It took me quite a while to figure out how to distribute addresses in my network and even get my pppd configured to request the prefix.

Re: 16 Years with IPv6

#46
Are these truly all the findings of the author after 16 years of running IPv6?

I find the list/number of points very little and most are common knowledge or been in countless of how to roll out ipv6 guides.

The mentioning that if end users have problems accessing your site, you should roll out ipv6 to your internal network, reads a 'bit' strange.

Re: 16 Years with IPv6

#47
post #44

IPv6 is not a requirement in all situations. At most enterprises need IPv6 at their edge to avoid traffic flowing through carrier grade nats. Internally private IPv4, plus nats for egress are generally fine for most use cases.

Internal IPv6 is much cleaner though, and less prone to failures (e.g. anyone VPNing in from a coffee shop that uses the same private IPv4 range will have a bad time). If you've already gone to the effort of getting IPv6 to the edge, surely at that point it's worth pushing it all the way through.

Not just that, also in the case when a big company acquires/merges with another big company. The suggestion to have loads of servers on a private range such as e.g. 10.x.x.x is nice as it should be unique.. until this assumption breaks due to combining two networks together. It'll result in conflicts, discovery that lots of places had the IP address instead of a hostname, etc. It seems much saner to use IPv6.

Re: 16 Years with IPv6

#48
post #29

Earlier quoted context omitted.

EIPs only cost money if they're unattached. Otherwise they're free.

That's what I am saying, the cost of an unattached EIP has not gone up has it?

I don't think so, but I do remember when unattached EIPs were free before they became not-free.

Re: 16 Years with IPv6

#49

I think of IPv6 like driving a stick shift car that you skip a gear with. If you're lucky, the engine won't stall but even if it doesn't, you'll be accelerating very slowly until the RPMs catch back up to where they should be. In IPv6's case they skipped 3 gears because they knew the car couldn't possibly stall, but here we are, barely accelerating after all this time.

Hm, this analogy is awkward. Definitely agree that it's taken a lot longer to become standard than I would like. I was discussing this with a coworker today, and we were reflecting on a similar technical rollout that really did take of. The various ascii encodings -> UTF8. This was a big shift at the time and took a lot of time to fix. In fact I know of one large DB that, after a huge amount of outreach with customer…

> but in general it is not solving a problem most people have (yet).

I don't think that that's actually true. People have massive problems due to NAT and overlapping address ranges and stuff ... but it's just commonly accepted as the way IP works, people don't realize that they could just deploy IPv6 and do away with all those problems.

Post reply on HN