A Selection of Interesting Papers and Documents

• Piège dans le Cyberespace (in French), by Roberto Di Cosmo (local copy in PostScript, see also more information). A lucid and witty analysis of the restrictions on individual freedom and the information tax imposed on unwitting users by Microsoft. The inevitable conclusion is that it is not in the user's nor in Europe's interest to use Microsoft products.
• Ode à la Bureaucratie (in French), by Peter Van Roy. A (still incomplete) diagram of the administrative steps necessary to import a French-bought car into Belgium. Most of the steps show classic bureaucrat behavior (e.g., hiding behind regulations, irrelevant complexification, mistakes). Few bureaucracies are as unnecessarily complex as in Belgium. This is partly due to the national character (competence is irrelevant, relations are everything; add as much complexity as possible a priori), but also to the nontrivial overlapping competences between federal, regional, community, provincial, and municipal governments. The resulting large disgressionary behavior by bureaucrats leads to gross inefficiencies and power games. Furthermore, nominations are often based on political favors rather than competence.
• The Cathedral and the Bazaar, by Eric Raymond. A study of the Linux software development model that was a key factor in Netscape's decision to release the source code of Navigator 5.0.
• How to Deconstruct Almost Anything, by Chip Morningstar. A lucid exposition of the concept of ``deconstruction'', which is usually hidden in dense literary fog.
• The OS at the End of the Universe. Douglas Adams presents a witty but sobering look at the limits of human credulity and one of Microsoft's many marketing blitzes.
• Twelve Ways to Fool the Masses When Giving Performance Results on Parallel Computers, by David H. Bailey. This humorous article outlines twelve ways to artificially boost performance rates of parallel scientific programs.
• The Mercury project. Mercury is a pure logic programming language with execution efficiency comparable to C. By adding type and mode declarations to a Prolog-like language and providing declarative equivalents for side-effect operations, Mercury achieves a better separation of logic and control than Prolog. Mercury gives up some of Prolog's dynamicity (e.g., it has no constraints, no unbound variables, and no assert/retract) in exchange for higher performance and arguably better compile-time error detection. Mercury is an example of putting religious fervor over practical usefulness: the language is even more restricted than Prolog, but this does not seem to bother logic programmers.
• High-Performance Prolog Implementation (1000K), by Andrew Taylor. Ph.D. dissertation, Basser Department of Computer Science, University of Sydney, June 1991. This dissertation constructs and measures an extremely efficient Prolog compiler, Parma, that generates code comparable to C in efficiency. This dissertation and Van Roy's dissertation give the final death-blow to the myth that Prolog is slow.
• Abstract Interpretation for the Compile-Time Analysis of Logic Programs (1200K), by Thomas Walter Getzinger. Ph.D. dissertation, Technical Report ACAL-TR-93-09, USC, September 1993. This dissertation is probably the most comprehensive taxonomy of analysis domains yet done that measures their usefulness when integrated into a Prolog compiler.
• The Costs and Benefits of Abstract Interpretation-Driven Prolog Optimization (180K), by Thomas W. Getzinger. Static Analysis Symposium (SAS 94), 1994. This paper provides the essential information of the above dissertation in a more compact form.
• Automatic Design of Computer Instruction Sets (1400K), by Bruce Kester Holmer. Ph.D. dissertation, UC Berkeley, 1993. This dissertation is an excellent example showing the usefulness of constrained exhaustive search. Automatic exploration can achieve results comparable to hand-designed instruction sets. A case study is done of a RISC specialized to Prolog, which is compared to the VLSI-BAM processor.
• Operational Semantics of Constraint Logic Programs with Coroutining (220K), by Andreas Podelski and Gert Smolka. A compact and easy-to-understand semantics that explains in a clean way both coroutining and cut.
• A Two-level Description of Stroopwafels (60K), seminal result by Christer Samuelsson.
• Members of DEC PRL in June 1994 (860K).