Domains with underscores are considered as invalid

[rozhn0v]

Hello! The library fails to parse domains with underscores.

let list = List::fetch().unwrap();
let domain = list.parse_domain("_jabber._tcp.gmail.com")
    .expect("underscores in labels are valid");

[eggyal]

This appears to be what List::parse_dns_name was introduced for, however if the root contains an underscore (or any other byte that's valid in a DNS label but not in a hostname) then the following check causes DnsName::domain to return None which surely is not correct?

publicsuffix/src/lib.rs

Line 537 in 149e9fd

if Domain::has_valid_syntax(&root) {

Can we remove this check, please?

[eggyal]

Actually, per #7 (comment):

Some people, myself included, are using this library to validate "registrable" domain names. As such, domain.root() must only return registrable domain names.

Nothing in the standards prevent registries from offering for registration domains containing bytes that are disallowed in hostnames, though I accept as a matter of policy most if not all do impose such a restriction. So whether this check strictly belongs in this library is not so clear.

However, if it is to remain, could we instead add root and suffix methods to DnsName that bypass these checks?

[rushmorem]

Can we remove this check, please?

@eggyal psl sounds like a better fit for your use case. It's lower level and only checks to see if the input ends with a public suffix. It doesn't try to validate the syntax of the hostname. As a bonus it's much faster, doesn't have any dependencies and it is no_std.

[rushmorem]

@eggyal Does this API work well for your use case? It should eliminate any false negatives you were having but your feedback here is highly appreciated.

[eggyal]

I've taken a quick look and indeed it does appear to fit the bill. Thanks @rushmorem !

[rushmorem]

Awesome! Thanks for the feedback.

[rushmorem]

I believe this is now fixed, both in this crate and in addr where parse_dns_name was moved to.