[beardygo]

Indeed. As a speaker of several languages, including RTL language (they haven't even considered the problems with RTL marks etc), I say stay with ASCII for usernames, keep UTF for full names.

If restricted ASCII a-z is good enough for passport names worldwide, it's good enough for usernames.

[macbr]

I’m confused – my name as written on my passport definitely contains non ASCII characters?

[extraduder_ire]

What is it in the machine-readable section at the bottom? My passport takes the apostrophe out of my name down there.

[belorn]

What is the point of a machine-readable name when there is a machine-readable passport number which should be unique for each issuing country? In this age I would assume that places which uses machines to read passports also are connected to international databases where the unique number is checked for validation. My country also mandated passport with chips in them for the last couple of decades, so by now there are no longer any valid passports without such chip.

If I had to guess, it seems the machine-readable section is just backward compatibility for machines built during the period where people started doing machine reading of passports but had yet to started to put chips into them.

(as a fun side note, smart phones can read the chip on passports and this is then used by some digital identity providers to establish identity on account creation, in combination with the phone camera).

[Muromec]

There is no database to query unless you issued the document (except revocation database). There is a chip with CMS signed data in it and MRZ is used for key agreement to read the data.

To know that MRZ and data arent from a different person or document, they have the name in ascii. It all kinda works and mskes sense in the end.

When you read the card with phone camera it uses mrz too

[belorn]

Looking it up, the mrz are only there to validate that the information stored on the document is the same as the information provided by the chip, and to make any eavesdrop attacks between the reader and the chip less likely to succeed. Its an optional standard.

The data on the chip is authenticated through a country signing key. This part is mandatory and prevent the person who carries the document from falsifying the information on the chip. There is also an optional active authentication chip to prevent someone from copying a passport even if they copy of the mrz and a copy of the traffic between chip and reader.

The MRZ is also part of the older standard which is intended to be replaced by a newer system that has card access numbers, which mean that the mrz and the ascii it embeds could very well be gone from passports. This new standard was implemented in EU by 2014, so there might passports issues now without the MRZ.

[macbr]

Oh, yeah. No non-ASCII in the “machine readable” part. Though I’ve never seen anything use that section. My national id card also has a “machine readable” section – but that doesn’t even contain my whole name: It’s just cut off after 20 letters.

[Muromec]

You probably have ASCII-adjacent name to begin with, so people who can read some kind of language using Latin letters will simply ignore "funny dots and dashes" and pronounce it kinda wrong.

It's on a different level from having a name originally written in a different alphabet entirely. At this point you just have it written in two scripts, with second being ASCII.

[mschuster91]

> If restricted ASCII a-z is good enough for passport names worldwide, it's good enough for usernames.

Passports (and credit cards) are the best example why ASCII-only is horribly broken. It's 2024, people want to type in their name as they write it normally, and they have the reasonable expectation of IT "dealing with it" behind the scenes.

Unfortunately, that expectation isn't reality, and it's all too common people are being rejected at the border or their card transactions are denied because braindead policies leave no other option but to blanket deny in case of mismatches.