silverwind

Not going to use this as long as node does not support it.

For reference, https://github.com/vitejs/vite/pull/12084 is the actual fix for this, need to install that handler.

> Based on https://datatracker.ietf.org/doc/html/rfc3339#section-5.6 i see `partial-time` and time could be without timezone, so v3 is doing it too much restrictive. Only way is to revert into v2 `partial-time` is...

I've seen these in use but think they are rather confusing. I'd suggest raising this issue first on [ip-regex](https://github.com/sindresorhus/ip-regex) first, maybe they want to support it with an option which...

> short notation such as 65.0/14 only applies to CIDR It's a entirely user-defined convention (that to my knowledge was never standardized) but I can't see why it should not...

I'm not against this, but first the change has to be accepted in [ip-regex](https://github.com/sindresorhus/ip-regex) first, maybe with an option, because we outsource that part of the regex from there.

Definitely a bug. The smallest numbers I could reproduce with is `merge(["::2:0:0/128", "::1:0:0/128"])`. At least safeguarding for division by zero does not fix it because the `subparts` function would recurse...

I added a workaround that pre-sorts the networks in `merge` which does seem to resolve it. It's definitely not the right fix as `merge` should work with any sorting, but...

Besides putting a caching resolver in front, you could use a in-memory cache and expire it after a fixed amount of time: https://github.com/medikoo/memoizee#expire-cache-after-given-period-of-time I'm not totally opposed to having a...

Hmm I think it's better to introduce a `opts.resolver` option on the public APIs `lookup` and `batch`. Exporting this "dirty" internal function seems wrong to me.