3 ms·
One of the issue with ascii domains is the phonetic mapping for CJK languages. For instance right not asahi.com is taken by 朝日 (Asahi shibun, the newspaper), t
by makeitdouble 3y ago
One of the issue with ascii domains is the phonetic mapping for CJK languages.
For instance right not asahi.com is taken by 朝日 (Asahi shibun, the newspaper), thus the Asahi town (旭) cannot use it. Mind you, the town could take asahi-town.co.jp or something like that, but there is other Asahi towns and places with different writings (including 旭日, 朝陽、浅緋 etc.)
Wanting all of them to have some random ascii diversification is madness, and we're only talking about Japanese places, when the Chinese character space overlaps. (The question of whether these domains are actually registered is I think a chicken and egg problem, and I have a hard time imagining the above conflict space will get a nice resolution inside the ascii alphabet)
Western ascii domains being ripe for scam is an issue, throwing the baby with the bath water would still be problematic.
- GoblinSlayer 3y agoAmbiguity exists in all languages, e.g. what mail.com refers to?
- makeitdouble 3y agoThere isn't as much ambiguity in the cjk writing though. It would be similar to mail.com, mel.com, meil.com, mehl.com, melle.com and all other variations that sound roughly the same getting all mapped to meɪl.com We could live in a world where all words are written phonetically, but we don't. Expanding that courtesy to non alphabetical languages would be a pragmatic and sane approach.
- necovek 3y agoThat seems to be mostly accidental and unrelated to ASCII: there are multiple towns named "London" and by the way that DNS works, only one of them can be "london.com" or "london.info" or... While I can see an argument for expecting to find whatever brand you think most relevant at BRAND.com(mercial), first-come-first-served is as fair as we can realistically get to (with some protections against misrepresentations and squatting). Note that I am not against IDNs: I just think your argument is flawed in their support. Again, Unicode is what allows all languages to be written out, and the fact similar-looking-URL issue is more present (it's present with ASCII too) is not a slight against Unicode and IDNs: we just need to solve for it.