Talk:But Is It Art?

This language fails to be IO-complete in a, perhaps, interesting way: Whenever (I,O) is a valid input/output pair, then so is (I^k,O^k) for any k > 0. So, for example, one can write a BIIA program that accepts strings of composite length, but no program that accepts string of prime length. Obviously this is not a very severe restriction; one can simply add an end-of-input marker to avoid this issue. But perhaps it would still be worthwhile to restrict witness rectangles to those that cannot be decomposed into smaller rectangles? Int-e (talk) 17:26, 12 April 2017 (UTC)

I realised this myself while offline. I'm leaning towards just leaving it, because as you say, it's at least a really interesting way to fail IO-completeness, and the fix in actual programming is trivial. I was leaning towards adding an explicit EOF marker (and maybe a corresponding BOF marker) as a means of fixing it, if necessary; it might make programming a lot easier. That said, the way the language can build up very complex constructs out of its current abstract input is somewhat appealing, and adding explicit BOF/EOF would mar that somewhat, so perhaps disallowing rectangles which contain other rectangles as subsets is a better fix if a fix is desired. --ais523 20:50, 12 April 2017 (UTC)