1 Introduction

Although the programming language LISP was first formulated in 1960 [7], a widely accepted standard has never appeared. As a result, various dialects of LISP were produced [1261089] in some cases several on the same machine! Consequently, a user often faces considerable difficulty in moving programs from one system to another. In addition, it is difficult to write and use programs which depend on the structure of the source code such as translators, editors and cross-reference programs.

In 1969, a model for such a standard was produced [4] as part of a general effort to make a large LISP based algebraic manipulation program, REDUCE [5], as portable as possible. The goal of this work was to define a uniform subset of LISP 1.5 and its variants so that programs written in this subset could run on any reasonable LISP system.

In the intervening years, two deficiencies in the approach taken in Ref. [4] have emerged. First in order to be as general as possible, the specific semantics and values of several key functions were left undefined. Consequently, programs built on this subset could not make any assumptions about the form of the values of such functions. The second deficiency related to the proposed method of implementation of this language. The model considered in effect two versions of LISP on any given machine, namely Standard LISP and the LISP of the host machine (which we shall refer to as Target LISP). This meant that if any definition was stored in interpretive form, it would vary from implementation to implementation, and consequently one could not write programs in Standard LISP which needed to assume any knowledge about the structure of such forms. This deficiency became apparent during recent work on the development of a portable compiler for LISP [3]. Clearly a compiler has to know precisely the structure of its source code; we concluded that the appropriate source was Standard LISP and not Target LISP.

With these thoughts in mind we decided to attempt again a definition of Standard LISP. However, our approach this time is more aggressive. In this document we define a standard for a reasonably large subset of LISP with as precise as possible a statement about the semantics of each function. Secondly, we now require that the target machine interpreter be modified or written to support this standard, rather than mapping Standard LISP onto Target LISP as previously.

We have spent countless hours in discussion over many of the definitions given in this report. We have also drawn on the help and advice of a lot of friends whose names are given in the Acknowledgements. Wherever possible, we have used the definition of a function as given in the LISP 1.5 Programmer’s Manual [7] and have only deviated where we felt it desirable in the light of LISP programming experience since that time. In particular, we have given considerable thought to the question of variable bindings and the definition of the evaluator functions EVAL and APPLY. We have also abandoned the previous definition of LISP arrays in favor of the more accepted idea of a vector which most modern LISP systems support. These are the places where we have strayed furthest from the conventional definitions, but we feel that the consistency which results from our approach is worth the redefinition.

We have avoided entirely in this report problems which arise from environment passing, such as those represented by the FUNARG problem. We do not necessarily exclude these considerations from our standard, but in this report have decided to avoid the controversy which they create. The semantic differences between compiled and interpreted functions is the topic of another paper [3]. Only functions which affect the compiler in a general way make reference to it.

This document is not intended as an introduction to LISP rather it is assumed that the reader is already familiar with some version. The document is thus intended as an arbiter of the syntax and semantics of Standard LISP. However, since it is not intended as an implementation description, we deliberately leave unspecified many of the details on which an actual implementation depends. For example, while we assume the existence of a symbol table for atoms (the ”object list” in LISP terminology), we do not specify its structure, since conventional LISP programming does not require this information. Our ultimate goal, however, is to remedy this by defining an interpreter for Standard LISP which is sufficiently complete that its implementation on any given computer will be straightforward and precise. At that time, we shall produce an implementation level specification for Standard LISP which will extend the description of the primitive functions defined herein by introducing a new set of lower level primitive functions in which the structure of the symbol table, heap and so on may be defined.

The plan of this chapter is as follows. In Section 2 we describe the various data types used in Standard LISP. In Section 3, a description of all Standard LISP functions is presented, organized by type. These functions are defined in an RLISP syntax which is easier to read than LISP S-expressions. Section 4 describes global variables which control the operation of Standard LISP.

Services provided by