Papers

Seen through the eyes of an inveterate open source hacker, SML/NJ is unusual in that it was written and is maintained by academics for whom papers published in refereed journals are the primary product ("publish or perish!") and the software itself merely sort of excreted effluvium.

This means in particular that the internals documentation cannot be found in the source tarball, with rare fragmentary exceptions (don't miss the README and MAP files), but must rather be cobbled together from dozens of papers. Places to check include:

• The smlnj.org page, in particular the SML/NJ-C foreign function interface and Compiler Notes links.

• The Flint project's publications page has many relevant papers, particularly toward the bottom.

• The MLRISC website is a few years behind the codebase, but has much good overview material.

• The home pages of the SML/NJ developers list their publications, many of which are relevant. Feeding their names into Google or (especially) Google Scholar is another way to turn up such papers.

  Be aware that papers are usually in pdf or postscript, so the contents are often not indexed in search engines. (If you find a given paper is listed as "only available to paid members" or such, feed the title back into Google -- with a little persistence you'll usually find a publicly available copy. If the paper has multiple authors, search under each, or check the home pages of each.)

• My Readings for late 2004 and early 2005 lists the papers I found most useful or interesting relative to SML/NJ as of that timeframe. Here are five good introductory papers which I recommend reading in the given order -- I would have saved a lot of time had someone given me that advice summer of 2004 when I started reading up:

  • Type Systems 1997 42p Luca Cardelli. (This explains the jargon you'll need to understand both other papers, and also the compiler source code.)

  • Reflections on Standard ML David B MacQueen
    
  • A Critique of Standard ML 1992 24p Andrew W Appel
    Overviews of what is right and wrong with SML, by the co-authors of SML/NJ.
    

  • Standard ML of New Jersey 1991 13p Andrew W Appel, David B MacQueen
    
  • Separate Compilation for Standard ML 1994 11p Andrew W Appel, David B MacQueen
    Together, these two provide a good high-level overview of the original SML/NJ compiler design and implementation.