Joseph Wright

I guess we didn't filter out the case of a 'no-variant' situation where the 'base' form is already incorrect.

> @josephwright Imho the message is correct. One should start with the Nn variant and then declare the cn and cV with `\cs_generate_variant:Nn ... {cn,cV}`. `cn` is a variant declaration...

> IMHO, considering N as a required base type is not sufficient because the 'c' to 'N' conversion need not be straightforward. For example, `\newcounter` conceptually takes a 'c' argument...

> Not really, a 'c' type argument gets expanded whereas an 'n' argument not. Nope: a `c`-type argument is expanded before passing to the base `N`-type parent, whereas `n`-type arguments...

See also https://github.com/latex3/latex3/issues/1038, which is likely to be affected by any implementation change.

Good question. I think we write _string representations_ of the data, and we say it's read back 'with current category codes'.

We discussed this briefly at today's team meeting: @blefloch has thought about it, but there's not an obvious way to solve it in general. It's still though on the to-do...

I looked at what I called 'data tables' a while ago. There are a few ways of implementing things, and I don't think I chose the best one. It's likely...

I've transferred to 2e as this is about the 'official' file

This is true in principle but at present we are not planning to change - the class is rather a hack at present. Leaving this open for when a refactor...