Hacker News Reader

Hacker News

Comment by karmanyaahm | Hacker News Reader

Full thread

karmanyaahm

imho ARIN should let these shady companies own IPv4. Only driving up prices of v4 will promote v6 adoption.

View on HN

If a government was smart, it would create an IPv4 tax and capture that value for itself.

Tax the inelastic and all.

For sure.

This has the secondary benefit of freeing up unused space.

Freeing up unused IPv4 space has the opposite effect, it delays adoption of IPv6 since it increases supply, thus lowering the price.

Which is fine. IPv6 is a pain in the neck to deal with. There is a fair bit of unused IPv4 space. Put a $5/year tax on IPv4 and a ton would free up.

This is unfair, for those waiting for ipv8.

Woh there, I hear that causes rabies in cancer cells.

Precog DNS is the future of subzero latency.

The US DoD owns 14 /8s and appears to be preparing to sell some of that space. Value captured.

Which country’s government and why?

CGNAT exists (and it works well).

Cancer exists too, and it's effective at what it does.

CGNAT provides positive value to humans. Do you have a point to make?

CGNAT does provide value in our current v4 situation, but it's far from ideal. Not having NATs allows for more decentralization (p2p connections, self-hosting), and saves the resources - unnecessary energy and compute - that NAT consumes

Dumpster fires exist (and are a valuable source of heating).

>CGNAT exists (and well... it "works").

FTFY. CGNAT means consumers get to be behind a double NAT, no port forwarding, poor P2P support, etc. CGNAT is the IPv4 solution of last resort.

IPv6 was poorly designed, and will never be fully adopted because of this. Unfortunately, we seem to be stuck between IPv4 and IPv6 when really, what we need is a new solution, something better designed and easier to adopt.

That's a strong statement about IPv6, is there an article or paper you are referring to?

Not really. Just personal experience. And I think that's the key point here - coming up with a design that is fit for the world (wide web), and not just academically viable.

I suppose the evidence I would give to back up my statement is that if it was well designed, it would have been adopted. As it stands, I only find it is supported among the more distinguished services that I use.

There's a difference between "poor design" and "difficult to adopt", and I think confounding the two doesn't make sense.

For another example, Python 3 has much better design than Python 2, and the difficult adoption a result of having people fix what they had written using the poor Python 2 design, and making it easier to adopt would have meant preserving the poor choices.

I don't know enough about IPv4 vs IPv6 to make a judgment in this particular case, but I don't think your argument broadly in itself here follows.

I've heard people complaining how hard IPv6 but I've never see a plan that is actually easier than migrating to IPv6. It's always massive CGNAT investment or making non-backwards compatible changes to IPv4 that is somehow "easier" to work with.

That is like all the how to terraform Mars plans. I know what is easier than living on any other planet, saving the one we are currently on right now.

That being said I think it is good people are so good at being non-conformist, at least if there was any other solution out there we would more than likely have found it by now.

The comparison is wrong. ipv6 is already fully developed, while ipv4 is getting crowded.

There are already more devices connected to the internet than there are ipv4 addresses. Already Android devices alone are more than 3 billion, PCs 1.5 billion. The limit is at 2^32 = 4.3 billion.

It would be a better comparison if Mars were already habitable and one only had to move humans over. Because that's all we have to do to make ipv6 suitable.

If I understand correctly, due to NAT we’re not limited by the number of computers vs available IP addresses, but rather by number of simultaneous connections vs (number of IPs) x (number of IP ports)

You are right, NAT is why the internet still works in the first place. However, NATs have many issues, like severely harming peer to peer capabilities. There is also additional complexity. My ISP frequently has downtimes of their NAT infrastructure.

Then there is the more complex routing tables. ipv4 is highly fragmented so users with many users don't have large contiguous blocks, but multiple smaller ones. This causes routing tables to bloat up, which increases routing overhead of the packets.

Lastly, there is the ridiculousness of paying for ipv4 addresses in the first place. Ideally, companies could just register them, having to pay a small service fee, that's it. In ipv6 this is the case, but in ipv4 you see Amazon paying huge sums for large ipv4 blocks. More importantly though, this cost is also handed down to customers, at least for hosters like Hetzner. For small servers, this can quickly be a significant fraction of the cost.

I guess I explained myself really poorly, but I think we mean the same here. To be clear what I mean is that somehow making ipv4 work is much harder than migrating to ipv6. And clearly terraforming Mars is insanely harder than terraforming earth.

Wouldn't this be the kind of thing that would be ripe for a startup to make easier/more accessible through abstraction? Maybe the risk of fracturing an already dense protocol into multiple competing approaches isn't worth it, but my personal experience with IPv6 is also that it can be intimidating and confusing for the uninitiated (of which I am).

IPv6 works fine. It's all over my house, my mobile devices use it all the time (they're probably IPv6-only over LTE at this point), and I don't have any problems when I occasionally have to type in an IPv6 address. I don't see how you'll wrench IPv6 out of the billions (?) of devices using it and make them use anything other than IPv4 at this point.

> IPv6 was poorly designed.

I'd like to point out some design decisions I love about IPv6:

- 128-bit addresses instead of 32-bit. I can't imagine having a shortage of routable IP addresses for a long, long time.

- I don't have to use NAT if I don't want because there are enough IP addresses to go around. We've grown so used to NAT that we take it for granted, not recognizing it for the ugly hack it is. On my home network, just about every machine has a routable IPv6 address.

- A consistent subnet: it's almost always gonna be a /64. When I lay out my networks, I don't have to regret specifying a /24 and then, as the office grows, wishing I could extend it to a /22. Every IPv6 subnet is big enough.

- A consistent subnet means that you don't need to use a subnet calculator as much.

- Not wasting .0 and .255 addresses (network & broadcast addresses). It's particularly bad with very small subnets (e.g. an IPv4 /30 wastes half its usable IP addresses).

- On that note, I like having my machines being able to have a .0 (::) address. My favorite IPv6 address? 2600::. You can ping it to see if your connectivity is working.

- ::1 is loopback. 127.0.0.1 is lookpback on IPv4, but 127 is a kinda weird number (okay, it's 2^7 - 1, but it's still weird).

- Without NAT I don't have to worry about network collisions when I VPN into corporate ("What? They're using 10.0.0.0/24? That's my home network's subnet! Dang!").

Reply on news.ycombinator.com