The work described here was completed in late 1983 but not published at that time. It attempts to provide a reasonably formal and precise definition of the Cedar programming language. The then current version of the language was perceived as an inadequate base for a number of planned extensions to the language and supporting environment; on the other hand, there was already a large body of Cedar code that could not simply be abandoned. These problems are dealt with by defining a small but powerful kernel language plus a mapping of existing Cedar constructs into that kernel. The kernel language introduces value spaces and operations over them that go well beyond what has been available in any implemented version of the Cedar language; it was to provide the basis for extension and simplification. The mapping from existing Cedar into the kernel provides not only a migration path for existing code but also a definitional method.

This report should be of interest to students of programming languages and their definitions. Most of the interesting ideas of the Cedar language appear in the kernel, which is described in Chapter 2. Such readers should note that the formalism used to describe the kernel has several known shortcomings. Its treatment of so-called dependent types is somewhat cavalier. A subsequent report by Burstall and Lampson [1] includes a more careful treatment of such types in a language very similar to the kernel. The present treatment also glosses over most of the definitional problems raised by the possibility of concurrent evaluation.

The report should also be of interest to Cedar programmers. Chapters 3 and 4 constitute the most complete, precise and accurate definition of the implemented Cedar langauge that has appeared to date. For a reader willing to make the effort to assimilate the concepts introduced in Chapter 2, this report can serve as an interim reference manual. The later chapters are painfully honest and complete; as the abstract notes, they say much more than anyone probably wants to know. As of March 1986, the only known differences between the description and implemention, other than minor bugs in each, are the following:

Both classes of readers should note that many parts of the kernel language have never been implemented in their full generality. Some of the current developers and users of the Cedar language would not even agree that the directions of evolution suggested by the kernel language are desirable or feasible. The claims about long-term goals and promised improvements in this report should therefore be taken as the personal opinions of the author.