> You seem to equate number objects with symbols whose printed name is
> the written representation of the number object.

Yes.

> This is a natural
> mapping at best.  Modifying it by adding a prefix of ":" yields just
> another natural mapping.  (The latter has the advantage that it
> doesn't invent something new, that it also works nicely with R6RS
> readers (which don't have the |...| syntax), and that for some eyes
> (at least mine) looks a lot more beautiful than vertical lines in
> parentheses.)
>
> I didn't say that using a number object like 123 in a library name is
> a new convention [*].  Mapping it to the symbol |123| would be.

People would very rarely need to type the vertical lines. Those don't
work in #!r6rs mode, and for new RnRS editions we can equate numbers in
library names with symbols, which means people can type unadorned numbers.

A merged R6RS+R7RS standard will already support both |...| and syntax
objects. You'll have to figure out what to do about numerical library
name parts anyway. You can't forbid them since R7RS doesn't, and they
have to have some interpretation that works with the syntax object
requirement you state.

> Yes, if at least one part in the full library name is an identifier,
> we have lexical context information. However, the ambiguity of which
> library name part to choose would then have to be resolved.  Always
> using the last would be more regular.

IMHO it's simplest to require the first library name part to be an
identifier. We could also look for the first or last identifier.