Paul Nkemdirim

I do the arithmetic by hand once, on paper, and then check it against a tool, because the arithmetic is what has to survive being scripted. Everything follows from that split: the network address is the address with the host bits cleared, the broadcast address is the one with them set, and the usable range is what sits between. The two addresses to subtract from the total are the ones people forget, and a subnet with a prefix of thirty or thirty-one has fewer usable addresses than the arithmetic suggests.

A subnet of thirty-one has exactly one, which is a valid entry and rarely what anyone intended. The classic subnetting question has an answer that depends on the prefix length being used, and asking for one number rather than a range is how a design goes wrong. For IPv6 the arithmetic is different in shape rather than harder.

The address is one hundred and twenty-eight bits, written as eight groups of hexadecimal with a compression rule, and a prefix of sixty-four is the convention. The subnetting question has a different answer because the address space is so much larger that a single allocation covers a site, which is the first thing worth saying to somebody porting a script.

I cover the tooling that reads the configuration, since a netmask that disagrees with the prefix produces a route table nobody can reason about. I keep the worked example visible so the arithmetic can be checked rather than trusted. I keep the worked example visible on the page so the arithmetic can be checked rather than taken on trust, which is the reason for doing it by hand once.