* complete the implementation of class I/O (see subroutines _read and
  _write in Language::INTERCAL::Object and TODOs in
  Language::INTERCAL::GenericIO)
* sick should use a standard (for INTERCAL values of "standard") grammar
  to parse its (mis)configuration file, say "sickrc.io"
* regular grimaces (INTERCAL's answer to regular expressions); currently,
  support for regular grimaces is included in the runtime compiler, but
  cannot be used until somebody writes the module. To do pattern matching,
  one would use a regular grimace in a CREATE statement, then try to
  execute a comment to trigger the new grammar production; this is probably
  quite unlike what other languages do. It should be added that it is
  already possible to do pattern matching using pseudo context-free
  grammars by using the CREATE statement without regular grimaces. You
  didn't really want to know that, did you?
* C version of the CLC-INTERCAL virtual machine, and corresponding "C"
  backend which generates C code to be compiled and linked to it. This
  is quite necessary because Language::INTERCAL::Object takes two seconds
  to load on my system, and is needed at runtime too (yes, I know about
  AutoLoader and SelfLoader; they actually make it slower... as for perlcc,
  it tends to run away screaming when it looks at these modules, I wonder why).
  The problem here is that each file in Language::INTERCAL:* depends on each
  other file, so (1) perlcc gets confused; (2) humans get confused too.

