Explain Like I'm Five

IPv6 has several changes to the specification, but since this is ELI5:

When you were a child, your friends would call your house and a parent would answer the phone. They'd ask to talk to you, and your parents would hand the phone off to you. That might have been because you were too young to have a phone, but IPv4 with NAT works the same way because there are so many "houses" and only enough phone numbers for the houses, not all the people that live in them.

For IPv6 it's like your friends can call you directly on your cell phone. And they can call your brothers and sisters, your cat, your dog, your TV, your refrigerator, and the backyard squirrels. There are so many phone numbers that everyone can have their own.

It's just a larger address space, nothing particularly special. Just more numbers

If you use IPv6, you don't need to have concepts like rfc1918 / "private ip spaces, so if every ip is unique, they can be used publicly you don't need NAT

I'll give it a shot. Not quite ELI5 but "Explain like I know what a phone number is". For the most important answer, see the last paragraph.

IP addresses are a bit like phone numbers. To send data to some computer, your computer attaches that number and sends the data packet on its way. With IPv4, an address is four bytes long, usually represented as four numbers from 0-255 separated with dots. That gives us a bit under 4.3 billion possible addresses which seemed enough when the system was invented and larger organizations could even reserve entire address ranges and some ranges got reserved for special purposes (for example, all 127.x.x.x addresses mean "send this to myself" while 192.168.x.x and 10.x.x.x are meant for local, non-public networks). Reserving these ranges is convinient when you need multiple machines connected to the internet but is very inefficient as these ranges need to be a power of two in size (256 is common), so you may get more addresses than you need and the rest stays unused.

The first solution was "Network Address Translation" (NAT). Basically, every household or organization gets a single public IPv4 address and every device on your network has a private address. On outgoing connections, your router replaces the (private) sender address with its public address and remembers which private address belongs to that connection so it can correctly forward any replies. For incoming connections, the router needs a list of rules to tell it what to do. For example something like "Everything on port 80 goes to 192.168.0.42". This worked for a while as most people make only outgoing connections and even many organizations can simply decide locally what to do with an incoming connection based on the received data so they wouldn't need multiple addresses.

After a while, it was clear that even with this workaround we would run out of addresses sooner or later. Providers tried giving their customers a different address every time they connected to the internet so they could reuse the address for someone else when the customer disconnected. This worked well when people only connected when they needed it but these days we're usually online 24/7.

So in the end, the only solution was to add more addresses. For our current needs, doubling the length would be more than enough but to be on the safe side, it was decided to quadruple the address length to a total of 16 bytes. This gives us about 340 undecillion unique addresses. Still not enough to give a unique address to every atom in the universe, not even enough for every atom on earth but still a lot. We can give every human an address range many times larger than the total address space of IPv4.

Does this mean that NAT is dead or that all your devices are visible from outside your network? Absolutely not. It means you can do that if you want. If your provider gives you a large address range, you can give each of your devices a different one and tell your router to forward everything. But you can also still use a single public address and/or tell your router to apply certain rules for what to do with incoming connections. There are also still address ranges that are meant purely for local use, equivalent to what 192.168.x.x and 10.x.x.x were in IPv4.

I'm going to just answer each point in turn. Maybe it's useful. I don't know.

It offers a shitload of IP addresses

It does. Generally most ISPs assign each user the equivalent of the IPv4 address space multiplied by itself. There's a lot of address space to go around.

They look really complicated

This is true. But you rarely need to remember a full IP address. Most resources you access via DNS. If you have servers on your own network you will probably need to remember your own prefix (first 3 or 4 blocks of 4 hex numbers) and your servers you want to access would likely be ::1 and ::2 etc in that allocation. So you'd learn them. Also most routers allow for local DNS entries and there's other things that will help here.

Something about every device in your local network being visible from everywhere?

This is a concern, but that's mostly because router makers now are often badly configuring their routers. The correct way to configure a router is to allow outgoing/established connections by default and block all incoming (until you specifically open a port). Once this is done the security is very similar to NAT.

Some claim it obsoletes NAT?

Yes, NAT was created to make a small address space work in an era of multiple internet consumers behind a single connection. But when each device can get a routable IPv6 address, NAT is not needed. However the security I talk about above IS essential to apply to consumer routers.

Now, I'll elaborate on some of the features of IPv6 (a lot of which are just not being used when they could have been).

IPv6 privacy extensions (RFC4951)

This allows normal client machines (the kind that would usually be behind NAT entirely) to have a similar level of security and privacy provided by NAT. One concern with just plain IPv6 with a fixed IPv6 allocation is that people could ID a specific machine from web logs etc and could be used against you in privacy terms. This extension ensures that you have multiple active IPv6 addresses. One could be the one you perhaps have some ports open on. That address will not be used for outgoing connections. A random IP will be used for outgoing connections and this IP will not have any ports open and will change frequently. I think on windows this is enabled by default (when you look in ipconfig you will often see multiple "temporary addresses").

Harder to portscan

Currently it doesn't take THAT long to portscan the whole IPv4 address space. And because almost every public address is hosting multiple hosts behind it, there's a good chance ports will be open on a lot of the IPs scanned.

With IPv6 the public address space is huge. With normal machines having their allocations made randomly within a huge allocation per user and every IP would still need every port scanned. This makes active port scanning much harder. The above privacy extensions also mean that passive port scanning (port scanning IPs found in web logs for example) is harder too.

User experience

Provided consumer routers are configured well from the factory and ISPs are making sensible decisions regarding allocation of address space, the user will benefit from the advantages and not even know they're using IPv6 in many cases. When you go to google/facebook/youtube etc you will be on IPv6 and not even know it.

You can set up your own IPv6 server, and it will be accessible from every other IPv6 device. Do you want to create a Lemmy server using only your phone and Verizon mobile data? Yes you can! *

* Instructions for installing Linux environment on Android phone and setting up free DNS and HTTPS certificates are not included.

That's the main point. IPv4 fails because half of all existing IPv4 devices cannot access the other half of IPv4 devices directly, they have to go around using some kind of relay server, which inevitably complicates communications and costs extra money.

I guess the thing I mostly don't get is:

If it's just an extension of the address space, then why is it still such a big deal?
Why didn't they add some sort of ipv4 compatibility layer into the new protocol and flip the switch everywhere 15 years ago?