Question

What libs have you used for that? How compatible are they with one another? Or did you write your own parsing routine?

I’m particularly interested in mutually-compatible implementations for Java, C++, Python, and JavaScript, which support:

• zero compression (“::“)
• IPv4-mapped addresses (“::ffff:123.45.67.89“)
• canonicalization (including to the short form, for human readability)
• CIDR-style netmasks (like “/64” at the end)

Answer 1

On POSIX systems you can use inet_pton and inet_ntop in combination to do canonicalization. You will still have to do your own CIDR parsing. Fortunately, I believe the only valid CIDR syntax for IPv6 is the /number_of_bits notation, so that’s fairly easy.

The other issue you will run into is the lack of support for interface specifications. For link-local addresses, you will see things like %eth0 on the end to specify what link they are local too. getaddrinfo will parse that but inet_pton won’t.

One strategy you could go for is using getaddrinfo to parse and inet_ntop to canonicalize.

getaddrinfo is available for Windows. inet_pton and inet_ntop aren’t. Fortunately, it isn’t too hard to write code to produce a canonical form IPv6 address. It will require two passes though because the rule for 0 compression is the biggest string of 0s that occurs first. Also IPv4 form (i.e. ::127.0.0.1) is only used for ::IPv4 or ::ffff:IPv4.

I have no Windows machine to test with, but from the documentation it appears that Python on Windows supports inet_pton and inet_ntop in its socket module.

Writing your own routine for producing a canonical form might not be a bad idea, since even if your canonical form isn’t the same as everybody else’s, as long as it’s valid other people can parse it. But I would under no circumstances write a routine of your own to parse IPv6 addresses.

My advice above is good for Python, C, and C++. I know little or nothing about how to solve this problem in Java or Javascript.

EDIT: I have been examining getaddrinfo and its counterpart, getnameinfo. These are in almost all ways better than inet_pton and inet_ntop. They are thread safe, and you can pass them options (AI_NUMERICHOST in getaddrinfo‘s case, and NI_NUMERCHOST in getnameinfo‘s case) to keep them from doing any kind of DNS queries. Their interface is a little complex and reminds me of an ugly Windows interface in some respects, but it’s fairly easy to figure out what options to pass to get what you want. I heartily recommend them both.