Peter.Lee@proof.ergo.cs.cmu.edu (02/23/90)
In article <1990Feb16.205321.5296@esegue.segue.boston.ma.us> bacic@ncs.dnd.ca (Eugen Bacic) writes: >Michael K. Gschwind writes: > >>Has anybody lately heard something about the MESS compiler generator or >>does anybody have a mail address of either Peter Lee or Uwe Pleban? Since the publication of my book, I have received over 60 requests for information about MESS. (Most requests come by email, some by phone or physical mail.) I try to reply to each query (in fact I have a standard form letter reply now) -- please forgive me if your query was somehow ignored. >The book by Mr. Lee is excellent and worth the read, but with the >sources being locked up somewhere it may be better to contact one of >Robert W. Gray, Vincent P. Heuring, Steven P. Krane, Anthony M. Sloane >or William M. Waite at compiler@boulder.colorado.edu (303) 492-5043. >They have an interesting product called ELI: a compiler construction >system. Thank you for the kind word about my book. The MESS system described in my book was developed as part of my Ph.D. thesis work. Though written mostly in Scheme, substantial portions were written in Pascal and Prolog. The system is not very portable, and there are some ugly bugs. The bottom line is that I have been too embarrassed to distribute this version of MESS, except to a few close colleagues. Uwe Pleban and I have been planning a new "Son of MESS" for several years now. This system will be an extension of the Standard ML of New Jersey compiler, and when finished we hope to distribute it. Until this new MESS is available, I think that your suggestion about looking into ELI (or comparable systems) is a good one. It is important to note that there is not really anything "magical" in MESS. MESS is simply a syntax-directed-transducer generator which uses Standard ML as the semantic metalanguage. Of course, MESS also provides a number of nice user-interface features, such as domain checking between phases, support for table-driven code generation, etc. But what is really important (and the point I tried to make in my book) is that such systems should be used with *high-level semantic descriptions*, which are split into what I call "microsemantics" and "macrosemantics". When this is done, I claim that "realistic" compilers can be automatically generated. This methodology can be used with ELI and many other systems. The only advantage that MESS would have over these systems is that it *enforces* the proper use of high-level semantics. >ps: Maybe this'll rouse either Mr. Lee or Mr. Pleban to answer, eh? ;-) Yup -- you got me... Peter Lee School of Computer Science Carnegie Mellon University Pittsburgh, PA 15213 Internet: Peter.Lee@cs.cmu.edu Phone: (412) 268-3049 -- Send compilers articles to compilers@esegue.segue.boston.ma.us {spdcc | ima | lotus}!esegue. Meta-mail to compilers-request@esegue. Please send responses to the author of the message, not the poster.