Champagne anybody?

[Patrick M. Hausen]

Good grief, watch Benedikt's talk, already.

I am a computer scientist, a software developer, a network professional, a FreeBSD and OPNsense contributor and don't you question my qualification.

I belong to the people who built this Internet as a mass medium thing. My company was founded in 1996.

[bimbar]

64 bits is not nearly enough for route aggregation and hierarchies.

I dare you to demonstrate that.
If you are not a software developer, it is hard to appreciate the true magnitude of a 64-bit number.
It gives you another whole range of IPv4 on top of itself. There is no way in hell this is not enough.

It seems you have never seen a structured IPv6 addressing scheme.

Which is, by the way, so much better than the "just use anything that's available" method of ip structuring we use with IPv4.

[chemlud]

Can we summarize the discussion:
ipv6 is perfect, just the users are to dull to realize it?

A solution to the problem "smartphone", nothing else. Hmm, imho it's not worth the trouble.

[Patrick M. Hausen]

Can we summarize the discussion:
ipv6 is perfect, just the users are to dull to realize it?

I am aware that various ISPs deliver dysfunctional IPv6 implementations violating official RIR recommendations. But claiming that the product is not necessary and IPv4 is all we will ever need is equally not helpful.

Official standards bodies:

- every consumer should get a static /56
- every small business should get a static /48
- every enterprise should get as much as they need

ISPs:

Now, how can we continue to ship "dynamic IP addresses" to force people to upgrade to the "business plan"?

As I said: morons.

[verfluchten]

We had 4 billion IPv4s minus blocks of private networks (a handful compared to the whole range) and supposedly ran out of them by 1998 when I spoke profusely against the new standard's wasteful bitness.

Now we have backwoods ISPs who are assigned blocks of 2k IPv4s and N*10^24 IPv6. I kid you not. But if we keep going like this, we will run out of even the 128-bit integer.

The fully-routable 32-bit IPv4 yields 4 billion addresses.
By adding only 1 extra byte, we would have gotten 1 trillion. Think of it!
By adding 2 extra bytes, we would have gotten 281 trillion. That would have been enough to give a routable IP address to every grain of rice on the planet. Imagine trackable grains of rice!
I'll skip 3.
By bumping IPv4 by 4 bytes, we would have gotten 18.5 pentillions-quintillions of addresses, enough to assign one to every electron.

All of that would have still been memorable to a decreasing degree and easily computable on relatively modern hardware. But it was not good enough for cowboys, and we settled on a 128-bit integer that no microcontroller or SOC can compute on silicon and needs libraries for.

Try to put yourself in a sysadmin's/sysop's/dev's shoes and to memorize a few dozens of this 100.101.102.103.104.105 or this: 2001:0000:130F:0000:0000:09C0:876A:130B.

Nuff said.

[Monviech]

dead:beef::1/64

2001:fedb::192:168:1:1/64

you can be creative xD

[bimbar]

Now I have read the original article (https://blog.apnic.net/2024/10/22/the-ipv6-transition/), it's actually quite interesting, but also, I think, shortsighted. I might also add that the word "obsolete" does not feature in this blogpost.

The central argument is that IPv4 address sharing (in the form of NAT and name-based virtual hosting) works quite well, and that the internet is not the peer-to-peer network we once thought, but mainly a method of delivering content from a central point to customers at the network edge.

However, I think that that point of view is already being made obsolete. The only observation is that we have learned to live with the dysfunctions of IPv4, and that the coping mechanisms make IPv6 unnecessary.
I would add that this heavily depends on the customer not being able to fully participate in the network, partly because he doesn't get a public address, and partly because he most likely has a internet connection that is too slow, and most of all, asymmetric.
But that factor is going away with the accelerating rollout of fiber connections to the customer. Another point is that the migration to the cloud does not seem to be as certain as it once was, with the prices as they are.

So in short, IPv4 is the communication protocol of a centralized internet optimized for the content consumer, while IPv6 enables the original decentralized approach that becomes more viable with the now available, much faster and symmetric, consumer internet connections.

[Greg_E]

Patrick,

I'd be very interested in a good long post about all the stuff that needs to be studied, maybe I can find a way around my CGNAT ISP, if they are doing IPv6 correctly. I think they are still handing out local addresses, but having a static internet routable address would be really nice.

[proctor]

https://www.heise.de/news/APNIC-Chefwissenschaftler-IPv6-Einfuehrung-wohl-obsolet-9995140.html

https://www.heise.de/meinung/Kommentar-IPv6-obsolet-Ich-glaub-ich-steh-im-Wald-9999396.html

[chemlud]

Auja! Wir verlinken hier die schönsten Kommentare vom Heise-Forum! Dafür sollten wir ein eigens Board einrichten! :-D

[Patrick M. Hausen]

I'd be very interested in a good long post about all the stuff that needs to be studied, maybe I can find a way around my CGNAT ISP, if they are doing IPv6 correctly. I think they are still handing out local addresses, but having a static internet routable address would be really nice.

Benedikt has got a number of great videos - he is one of the leading experts in Germany:

https://www.youtube.com/@benediktstockebrand9719/videos

Click on "oldest first" to see mainly IPv6 related content.

Also Ivan Pepelnjak's blog is great, although you have to search a bit for IPv6 related things. Ivan is one of the most prominent people making the case that ULA is useless in actual deployments.

https://www.ipspace.net/Main_Page

HTH,
Patrick

[Monviech]

ULA is useless in actual deployments.

https://www.youtube.com/watch?v=oAp6RUNEN-U

[proctor]

Auja! Wir verlinken hier die schönsten Kommentare vom Heise-Forum! Dafür sollten wir ein eigens Board einrichten! :-D

Heise-Forum?

[bimbar]

Also Ivan Pepelnjak's blog is great, although you have to search a bit for IPv6 related things. Ivan is one of the most prominent people making the case that ULA is useless in actual deployments.

https://www.ipspace.net/Main_Page

HTH,
Patrick

Especially his small site multihoming with ipv6 articles are very topical.