I don't know what's most common these days, but quite to the contrary there's often a great deal of agitation for full i18n of name fields, such that people can specify their preferred name / spelling, in their native language. This is often pitched as a rights/respect issue.
I feel otherwise: usernames and other identifiers (URLs particularly, and a number of other widely used identifiers), should be drawn from well-known, common, and small charactersets. The original 7-bit ASCII characterset (128 distinct characters), or some subset of that (e.g., Roman alphabet, Arabic numerals) may not be the native representation of all cultures, but it is all but certainly a familiar representation, and avoids numerous homoglyphic attacks.
This greatly exceeds the typical vocabulary (distinct known words) of most native-language speakers, estimated at between 20k -- 35k words for a typical adult English speaker.
Yes and pretty much all languages already have widely used and well known ASCII embeddings aka romanization schemes because early computing did have many places where more wasn't possible or at least not for more than one language at the same time.
I feel otherwise: usernames and other identifiers (URLs particularly, and a number of other widely used identifiers), should be drawn from well-known, common, and small charactersets. The original 7-bit ASCII characterset (128 distinct characters), or some subset of that (e.g., Roman alphabet, Arabic numerals) may not be the native representation of all cultures, but it is all but certainly a familiar representation, and avoids numerous homoglyphic attacks.
<https://en.wikipedia.org/wiki/IDN_homograph_attack>
By contrast, Unicode has something on the order of 160,000 distinct characters.
<https://en.wikipedia.org/wiki/Unicode>
This greatly exceeds the typical vocabulary (distinct known words) of most native-language speakers, estimated at between 20k -- 35k words for a typical adult English speaker.
<https://wordcounter.io/blog/how-many-words-does-the-average-...>