[sph]

So you migrate to int64 and one day someone will wonder why the hell did we ever think no one would reach 2^64 rows in a database table. Or that 2^128 IP addresses would be enough for everyone.

[WhitneyLand]

“32-bit IPv4 addresses to 128-bit IPv6 addresses means going from 4.3 billion to 340,282,366,920,938,463,463,374,607,431,768,211,456”

Yes 128 bits is only 4x times the bit length, but the address space is exponentially bigger.

Some predict if we do run out, it might take ~100 years.

[josephg]

> Some predict if we do run out, it might take ~100 years.

You think we can use 340,282,366,920,938,463,463,374,607,431,768,211,456 addresses in 100 years? Thats ... ahem, ambitious!

[Dylan16807]

In practical terms, IPv6 is arranged as a 64 bit subnet ID and 64 bit device ID.

And the subnets are in a hierarchy that helps with routing but limits the packing efficiency.

We could feasibly "run out" to the point that we can't keep doing things this way.

[ipython]

The truth is that the address space is extremely sparse. Every ipv6 subnet is a /64 by default, yet I’m sure most of us have vastly fewer than 2^64 machines in our networks … at least for now

[sokoloff]

In my experience, you first do an emergency re-seed to use the negative half of the int32 space.

Been through it twice in my career…

[marcosdumay]

Those numbers are much, much larger than 2^32.

No process that touches the real world gets into 2^64. No process at all gets into 2^128, at lwast not while we live around a single star.