[marwis]

Why not combine zero width character with visible character, i.e. use 2 characters for separators?

,<FS> for fields \n<RS> for records

This removes ambiguity in parsing and remains user readable. It's also relatively easy to auto-fix files edited by users in normal editors.

It also mostly removes need for escaping.

It's also smaller or same size as unicode multibyte characters (haven't checked).

[Renaud]

Wouldn't this easily break your file in subtle ways when someone tries to edit it in their editor and the zero width character is not visible?

How could you make the difference with a standard CSV file if it looks like a standard CSV file?

They explain why they don't use control characters. Editors are not consistent in how they show control/zero-length characters:

https://github.com/SixArm/usv/tree/main/doc/faq#why-use-cont...