Why don’t Chinese web addresses just use Mandarin characters? Because that’s a pain, too. The Internet Corporation for Assigned Names and Numbers (ICANN), which sets the rules for web addresses globally, has periodically hyped the expansion of domain names to include non-Latinate scripts, but Chinese web sites have yet to take full advantage. Some devices require a special plug-in to type in Chinese URLs, and even then it takes longer to type or write out characters than to input a few digits. Plus, for web sites that want to expand internationally but don’t want to alienate foreign audiences with unfamiliar characters, numbers are a decent compromise
I'm really wondering how many (or what percentage of) devices actually need "a special plug-in". Seems like it would be a very small - and shrinking - fraction, but I could be wrong.
If the user's device is (presumably) already set to use Mandarin (or any specific language), then I assume that would be the default input method, so why does that take significantly longer to use than numbers? I'm genuinely curious, never having experienced that use case myself as a native english speaker.
As for websites wanting to expand internationally and avoiding unfamiliar characters, that doesn't really make sense to me - if they want to expand internationally, I don't think the numeric domains are going to help much (they'd be just as cryptic to me as Mandarin characters), so why wouldn't they just register alternate domains in the target regions/languages as many already do?
Yes the meaning would be just as cryptic, but the act of typing and probably memorization would be much easier. As an exercise, try downloading an input method for a foreign character set and typing out a string that you see rendered in an image.
Many TLDs disallow registrations containing punycode. Take a look at Wikipedia's (incomplete?) list of those that do: https://en.wikipedia.org/wiki/List_of_cctlds
The one technical challenge I suppose could be a concern would be optionally implementing some sort of test to detect domain registration attempts that use similar-looking characters to those in existing legitimate domains (e.g., registrations from scammers attempting phishing attacks with domains that look similar to other domains but use different extended characters).
Still, it feels like the lack of support from these TLDs is more a matter of policy than a technical challenge.
This is called the IDN homograph attack (https://en.wikipedia.org/wiki/IDN_homograph_attack).
As I understand, punycode is translated by the browser to a standard ASCII string, so there's no need for special support in the DNS system or other infrastructure other than a simple translation in the browser's URL handling. Seems like a pretty straightforward/simple solution.