> Gauche had been using \xHH as literal octet value and \uHHHH as
> codepoint value in literal strings, before R6RS.  Now \xHH; is for
> codepoint value, so we need another escape sequence to literal octets if
> we are to use #u8"....".
> I'm wary of doing so, since it's yet another escape sequence that can
> only be usable in limited context.

It's good to keep in mind that we can always do (bytevector-append ...
(string->utf-8 "foo") ...) to accomplish anything that #u8"..." can do.
I imagine most fast Schemes can evaluate it at compile time. It's a
little more verbose, but blocking some escapes in #u8"..." doesn't mean
that people are blocked from making byte strings like that.

If we use #u8("foo") then the "foo" part looks exactly like an ordinary
string, so I sympathize with the argument that it should support the
same syntax as ordinary strings. However, hopefully having it inside the
#u8() is a clear enough sign that it's different.

To prevent problems, unrecognized \ escapes should raise an error in the
reader, not silently fall through. I don't remember how normal Scheme
strings are handled in this regard.