The time has come: we must enable IPv6 entirely

I see no reason, no reason whatsoever, to only give 2.pool.ntp.org an AAAA record. Let’s enable AAAA on the others too.

10 Likes

Just looking at the server count, there appears to be merit in enabling the use of IPv6 on another of the entries. Or maybe not. I will have to do some digging. Right now if everyone was on IPv6 I assume the clients would all prioritize IPv6 for each of the numeric entries, meaning that right now they’d talk to IPv6 servers 25% of the time. Adding one more AAAA record would bring that up to a 50/50 split.

             IPv4 IPv6
Africa         37 16
Asia          177 90
Europe       1680 982
North America 632 302
South America  45 12
Oceania        92 36
Global       2660 1426
1 Like

IOW, 35% (1426 / (1426 + 2660)) of the pool servers are on IPv6. Since the minimum number of servers configured is 3, the current setup, which provides AAAA records only for the 2.pool.ntp.org name, roughly matches the current distribution of IPv6 servers in the pool.

However, when using just the name pool.ntp.org, perhaps with the pool directive, no AAAA records are returned. Given the current fraction of IPv6 servers in the pool, it should return at least one AAAA record among the four, which would still provide three servers to IPv4 only clients.

Perhaps returning AAAA records for the name 3.pool.ntp.org could be considered too.

1 Like

Perhaps returning AAAA records for the name 3.pool.ntp.org could be considered too.

Great idea! How could we get that done?

1 Like

I have played with IPv6 but it’s a bad design, because if your NIC is replaced it gets a new IP and you can change all records again.

I tried to understand the stupid protocol but it makes no sense.

They should bury it and let people decide if they want NAT or not as today you can not use NAT and that means that every change in your local network, you have to do a lot of work to get things back to working order again.

I stopped trying to understand it, as it seems almost nobody understands it or can explain it.

Time for IPv4+, keep the current but just add a few more numbers, or change them from 256 to max 1024 and the problem is solved.
I never understood why they didn’t simply make more numbers and keep what we have.

IPv6 is stupid and it will die soon. Almost nobody uses it because of it’s dumb complex working.

1 Like

The NIC based IPv6 address is dynamically allocated. You can assign NIC independent IP address to the interface. Then, use that IP address in the pool.

2 Likes

Funny thing to say from someone who lives in a country with one of the highest IPv6-adoption grade in the world, where there is more IPv6-usage than IPv4 (source: IPv6 Capability Metrics).

:slight_smile:

In my case, I just don’t get the point to have more available addresses but not running it behind NAT like IPV4 does. For example, on my docker setup, I had to run a container that do NAT, to be able to run my pihole with IPV6 forwarders. Why docker is not able to run IPV6 in NAT mode like IPV4 does now… Maybe I’ve just not found the option yet.

IPV6 is different, we need to learn it like everybody does with IPV4 at the beginning. But it doesn’t make sense for me to say " every devices has a routable IPV6 address on Internet ". Local networks must be natted.

You may thinks so, but it’s not adopted at all.
Providers just add it standard to routers, but people don’t turn it off.

I know a lot of people here and not one of them is activly using IPv6, they all ignore it.

Exactly my problem too. I want NAT and just 1 public address at my router. But it won’t let me do it that way.
It’s just too complicated to manage and nobody seems to understand it except maybe the freaks that came up with the idea.

As much as I would love to go into discussions about it, this is not the place for a discussion about the usefulness or alleged novelty[1] of IPv6. There just one simple fact which is relevant; there is an increasing amount of IPv6 clients using the pool, wether some like it or not. And the pool should serve them, just as it serves IPv4 clients.

[1] IPv6 now exists longer than IPv4 existed when IPv6 was invented. Even though it might not be perfect, it’s not new, it’s not hard, and it is invented for a very good purpose. Contact me offline to learn all about it, I’m happy to help!

5 Likes

There aren’t enough IPv4-addresses to accomplish that on the longer term. Already people are being put behind Carrier Grade NAT, sharing scarce IPv4 addresses with an entire street or district. Try to imagine what happens if many IoT-devices attempt to synchronise at the top of every hour from an ISP that deploys CGNAT (mine does). They resolve pool.ntp.org and the ISP’s resolver hands all of them the same IPv4 addresses (and never any IPv6). That is a lot of rate limiting kicking in on those poor NTP-servers (who only see one single CGNAT source address sending a lot of NTP requests). Not very productive at all, right? Now try to imagine that this scenario is likely to be a reality already today and it’s not going to get less in the future. Quit the opposite.

3 Likes

You’re right about the fact that it’s not the place to discuss about IPV6 vs IPV4 but the point was about having 1 IPV6 address per house and people do what they want on their internal network.

1 Like

Marco you are missing the point.

If big tech like e.g. Microsoft, IBM and MIT wouldn’t be so greedy for address-ranges we wouldn’t have this problem and IPv4 has plenty space left. It’s just a few sites that own 2000000000 addresses, that is the real problem.

Goodluck with your IPv6, I won’t touch it anymore and stick with IPv4.

This was never about trying to convince you (or anyone) to appreciate and use IPv6 (although now you are missing out on https://42.dnslabs.nl/). It’s about adding more IPv6 DNS-records (AAAA records) to pool.ntp.org. And more IPv6-capable servers too, if at all possible. Simply to serve an ever increasing amount of IPv6-capable clients properly, rather than via any CGNAT addresses or anything similarly sub-optimal.

3 Likes

You wrote at the top: The time has come: we must enable IPv6 entirely

At the same time your name pops up at the kpn forum where you are pushing IPv6 in 2017.
There also nobody cares about IPv6.

You seem to be the only one :slight_smile:

300Gbps in Amsterdam on IPv6.

33% of Google users use IPv6.
https://www.google.com/intl/en/ipv6/statistics.html

I’ll let the numbers speak for themselves

If I get this right, this odd DNS setup to serve only AAAA records on one of the authoritative servers is yet another way to load balance traffic.

Otherwise, if all servers would also serve AAAA, it could potentially overload v6 severs while v4 servers would be just fine, given there are more IPv4 NTP servers than IPv6 servers.

That is indeed an odd use of DNS, but the pool is not a typical zone; so it makes sense in terms of using it to load balance, albeit it’s odd for DNS folks.

Does that make sense?

The numbers are speaking because a lot of devices on IoT are using IPV6 and we just don’t care about those devices, like my mobile phone that use IPV6 address when connected on LTE network. I just want connectivity, that’s all. Same for my VPS that use IPV6 too, it works, good ! I suppose that cases like that are the way to go. It’s just too complex if somebody is hosting an internal NTP server for the pool because IPV6 and NAT seems a PITA (for the moment).

Enough talk about that guys, I think there are enough arguments from each side. The project need good IPV6 servers and IPV4 servers. Time will write the rest of history…

1 Like

What about Android users? So you’re saying you don’t care about them when they’re on LTE?
I think this is rather unfair to them.
([0--3].android.pool.ntp.org)

Maybe it’s a given for you and other folks, but again, is this designed for loading balance v4/v6 traffic? (see my previous comment)

For the record: I am not yet advocating full support for v6 given that the v4 and v6 resources are heterogeneous. I’m just trying to understand first the design choices.

1 Like

I said device, not user. And I’m talking about my situation. If my mobile phone has an IPV6 when connected to my mobile data network, why i would care about the type of ip addressing, I’m not remotely connecting to that device anyway. So my mobile phone has an IPV6 address, connecting to the pool with IPV6 and get time from a server talking IPV6 too ? Good job ! We are saving some IPV4 adresses. No problem here no ? Why arguing on that, everything is fine.

My internal network is IPV4 and I just can’t easily get time from the pool with IPV6 (bad exemple here as my ISP not giving me an IPV6 address anyway). So no problem here too … IPV6 is bad for internal network or NAT. And no I don’t see any case where my internal NAS will be IPV6 because I don’t need that. My ISP has to give me an IPV6 address and I will NAT to it.

Topic : We must enable IPv6… do it. IPV4 must stay too (for the moment) that’s all.