So how is this website even possible?

[Rubbereon]

Hey so this website uses a japanese word for its top level domain name

https://panley.みんな/

I thought only ascii was allowed for a top level domain like com, net, org?

Can I do something similar for my website if I wanted to?

[ribose]

Can I do something similar for my website if I wanted to?

yes, you just register it like any other domain name. here are the ones available at porkbun (prices in USD)

[ThunderPerfectWitchcraft]

Its possible for quite a while now, see https://en.wikipedia.org/wiki/Internationalized_domain_name
As far as I can tell, the domain you found (".minna") is intended to be used for the japanese market, but you should still be able to buy a domain with it.

[Melooon]

There is a famous case where a company (Panic Inc) in 2011 figured out they could register the poo emoji as a domain name  - https://blog.panic.com/the-worlds-first-emoji-domain/ - it does not work for me in modern Firefox, but it does work in Safari!

Your domain name is actually still ascii, but most browsers will render it in unicode if they detect that its a unicode string.

[Rubbereon]

There is a famous case where a company (Panic Inc) in 2011 figured out they could register the poo emoji as a domain name  - https://blog.panic.com/the-worlds-first-emoji-domain/ - it does not work for me in modern Firefox, but it does work in Safari!

Your domain name is actually still ascii, but most browsers will render it in unicode if they detect that its a unicode string.

It's called punycode and it's not the same thing as a top level domain like .com, especially if it's being written in kanji. I dunno when they invented punycode, but I can assure you it works with anything and pretty much everything although it still follows the [anything].com formula.

How is it even possible for non-ascii characters in a top level domain? I know about percent encoding, but it's only valid for directories (example.com/[%put% %anything% %here%]) Doesn't ICANN verify if a domain is valid?

[Rubbereon]

yes, you just register it like any other domain name. here are the ones available at porkbun (prices in USD)

Actually your post solved the mystery, the TLDs are written in punycode too 

See this one? https://porkbun.com/tld/xn--q9jyb4c

The punycode appears at the end.

What's weird is that the kanji domain in my post displays the encoded kanji while @Melooon's poop emoji domain displays unencoded as punycode on librewolf. I guess there's a weird problem with gecko browsers like firefox that prevents them from always encoding punycode correctly.

[Melooon]

poop emoji domain displays unencoded as punycode on librewolf

Its actually explained in the article I linked (from what I can seee), you're not supposed to be allowed to use poo emojis in most URLs; the unicode that works is supposed to be on an approved list, and almost all approved lists don't include emojis. The poo one was allowed by mistake, and Safari is either respecting that legacy mistake (or prob just not enforcing anything), whereas open-source browsers like Firefox with nerdy-er development teams are enforcing the rule.

[Dan Q]

I guess there's a weird problem with gecko browsers like firefox that prevents them from always encoding punycode correctly.

It's actually a security feature. Early on in IDN rollout people realised they could buy e.g. рaypal.com which looks to the human eye to be the same as paypal.com but it's not: that first one used a Cyrillic 'р' as its first letter but the second one uses a conventional Latin 'p'. The last thing you want is for a domain name to look like a different domain name!

the unicode that works is supposed to be on an approved list, and almost all approved lists don't include emojis.

This is the solution to the security problem: an "allowlist" of characters that are considered unambiguous and safe for use in domain names. But because it's an allowlist, rather than a denylist, every character is assumed to be illegitimate unless/until it gets added to the list. And there are a lot of emoji. And some of them are pretty-similar to one another, especially across font-sets, so they'd be a challenge. Also for screenreaders! How can you tell by listening to your screenreader whether "poo emoji dot com" contains an emoji or is written out longhand? You can't; not easily. So it's ambiguous.

tl;dr: bad people doing bad things is, as always, why we can't have nice things.

[Rubbereon]

Gonna post this here since it's related to the conversation we're having: https://url.spec.whatwg.org/

I was trying to fact check something on the thread and found that website instead which also points to a unicode-to-ascii algorithm.

That might have been what Melooon was referring to in a reply earlier, although it converts the unicode text to either percent encoding or punycode depending on where the non-latin characters are used, it doesn't really decode anything.

Anyway what I was trying to fact check is the allowlist especially if it's inconsistent between Chrome and Firefox, but I can't find where it is discussed.