TODO

* Complete the bibliography

* Write a variant of libgc suitable for the linear lisp

* Write a variant of the core language interpreter,
   * in the style of maru? of gambit?

* Bootstrap a fuller Linear Lisp on top of the core language
   * in the style of gerbil?

* Use ptrace and/or /proc for editing memory in a foreign process
   * by default manipulate an external slave, don't modify current image
   * initial state for the external slave is a mindless worker waiting
   * still have special ways to work in the same process? atomically?
     By creating a slave that hooks back onto the detached parent?
  * Bibliography:
    http://dev.exherbo.org/~alip/pinktrace/
    http://www.linuxjournal.com/article/6100
    http://www.linuxjournal.com/node/6210
  * Use SYSV shared memory (or mmap to same file?) to share data with child.
    beware read/write barriers and icache flush.

* Reuse the insns.dat from nasm and write an assembler / disassembler
   * have a maru-style assembler, except defined based on the nasm database
   * use some reversible computing framework if possible
   * use a framework that makes partial evaluation easier
     so we can assemble faster by eliminating unused cases.
   * if that helps, start by implementing 64-bit mode only
   * General x86 information: http://www.sandpile.org/
   * Nasm: git://repo.or.cz/nasm.git

* Bootstrap with some maru-style emit.l based on the above assembler

* Maintain a content-addressed database of slave processes vs source
   * where "source" means "all the things that led to building that"
   * where "source" gives us complete offline debug information
   * where we also have intent-addressed mapping of source to object

* Profit!