As a bonus, the Acme::Bleach module will take an existing perl program and convert it (overwrite it) with a version written in whitespace.
"The first time you run a program under use Acme::Bleach, the module removes all the unsightly printable characters from your source file. The code continues to work exactly as it did before, but now it looks like this:"
http://search.cpan.org/~dconway/Acme-Bleach-1.13/lib/Acme/Bl...
Also relevant is Damian's Rod::Logic module. Unfortunately this hasn't reached CPAN (yet!) but examples can be found in his talks (for eg: http://blip.tv/open-source-developers-conference/temporally-...)
Error handling
If your program has a runtime error or a compiletime error the interpreter
will print "Parenthesis Mismatch" to standard output and then exit.
Such a set may consist from only one function. So, the whole program is a structure of calling that function. Writing down such a program, we obviously don't need to write the function - it comes by default; we only need to write the structure, which can be represented by ( and ) . In a different way, than what I see here, but still - one bit per parenthesis, very simple processor, lots of memory and lots of steps for simple operations - a Turing tarpit...
The first idea is that you normally use at least 2 symbols for a programming language, but you can in principle get away with just one. The "vanilla" form of this idea is to use the length of the processed string as the instruction itself -- that is, it has a bit pattern which gives you two symbols to play with in principle. Of course that's pretty ugly so you can instead do something slightly more "advanced" and use a non-character as a delimiter; the most obvious choice is file boundaries on a filesystem, but in principle any data structure boundary can be used.
The second useless language that I want to create is one which I'm not even completely sure works. The idea is that you have some number of nodes -- let's say 256 -- which each contain some number of bytes -- let's say 4. Those bytes represent two (node pointer, instruction) tuples. These get initialized to a definite 1KB prelude, and the computer starts at node 0. The instructions can also manipulate registers and a much larger memory space, and can load ints from that memory into other nodes, but the 1 KB is needed for the interpreter.
The idea for this useless language is that, the computer chooses one of the two tuples randomly, performs its instruction, and then goes to the node it points to. >:D
The basic idea is that all esoteric languages seem to be about "how do you program familiar thing X if you don't have Y?", but I have 256 instructions so I'll give you all sorts of wonderful functions to call on your registers. The problem is really just that the computer does not natively want to follow a deterministic path, and you have to "herd it" into that mentality.
[1]http://en.wikipedia.org/wiki/Esoteric_programming_language#T...
I initially thought it would have been a pain to write just the test programs, but I guess they were just programatically search-replaced from Lisp programs?