Issues [that I remember]:

1. It uses names parser and writer for the attributes. I don’t know if reader and writer would be better? parser and serializer? Can we please bikeshed this until the end times?!
2. The syntax for taking all the args into a single object abuses the C variadic signature syntax, and this kind of sucks (I would rather prefer it to be a spread-like syntax ...args: Ty), but maybe this sucks less than writing a custom version of syn::ItemFn? I will do the custom one if someone makes me, but I was feeling lazy every day and it was a very convenient already-existing syntax that I could abuse.
3. It sure is tempting to replace all of the custom trait impls’ fn read_options and fn write_options with this to create a single source of truth for what R is, since trait aliases still aren’t a thing. Should this be considered idiomatic ever?