> Do you mean this: #u8(1 2 3 4 5). [...]
>
> Would this be a terrible idea: #u8"0123456789abcdef". So it's a string
> but its contents must be an even number of hex digits.
>
> The trouble with that is splitting up long strings. Is this baroque:
>
> #u8{
> "000089f809d08b1d483e00004489e683"
> "ce0809d8448a15f13c0000410f45f445"
> "31c944380de73c0000410f45f44584d2"
> "410f45f44584c0410f45f48a05bb3c00"
> }

Or we could use the parentheses from R7RS:

#u8("000089f809d08b1d483e00004489e683"
     "ce0809d8448a15f13c0000410f45f445"
     "31c944380de73c0000410f45f44584d2"
     "410f45f44584c0410f45f48a05bb3c00")

Now it's just like R7RS but elements must be strings instead of
integers. (To be pedantic, they're not vector elements anymore, but
slices.) This is a break from RnRS but I quite like it.

We could also permit optional spaces in the strings. Is that useful? Hex
is hardly that readable to begin with.

Uppercase vs lowercase vs either for the hex digits? I don't care much.
Forcing a particular case is simpler and more uniform.

We could permit base64 bytevectors: #base64{"aGVsbG8gd29ybGQ="}