Behavior of incf with characters?

[Stephen J. Turnbull]

Ben Wing writes:

 > Checking the type of an object to determine if it's a character is not
 > expensive.  Implementing "next in charset" semantics is fairly cheap
 > if done in C and if we don't try to implement the "skip over holes"
 > behavior.  I'd tend to disfavor skipping over holes in any case --
 > from XEmacs's perspective these are valid characters, just with
 > undefined semantics,

I think this is taking the representation way to seriously, especially
since we all basically agree that the representation has to be
replaced with some sort of internal Unicode, where there *are*
non-characters (about a dozen in the BMP) that will never ever be
characters.

BTW, what is your use case for incf (or other arithmetic) on
characters?  I don't actually see a real use case for incf returning a
character, especially one of "undefined semantics".