[KevinH]

I like that idea!

But the encoding of the special_characters.txt file would be crucial. No utf-16 and bom, just pure utf-8 with no bom. There is also the problem of copying and pasting into the special_characters.txt file from other apps messing up the encoding, changing line ends, no newline in list line, etc.

In the worst case we could replace the character field with 2 to 6 hex chars, followed by a name, then self generate the actual character from it. You could use the same mechanism we use now to show a description if the character under the cursor in CV to generate a name description. You could also generate a numeric entity instead of a named entity from that code. Named entities require a special doctype under epub2 and except for the xml basic ones they are illegal in epub3, so their usefulness are about done.

I guess as an alternative we could create a special_characters.xml file that would have a clear utf-8 header, and that would come with better builtin parsing for more consistency.

I just fear that encoding things in text files under Windows text editors especially is going to be an issue for many users.