Dynamic scoping provides an important abstraction for dealing with side effects in a controlled way. A low-level procedure may have state variables which are not of interest to intermediate routines, but which must be controlled at a high level. Dynamic scoping allows any procedure to get access to parts of the state when necessary, but permits most procedures to ignore the existence of the state variables. The existence of many dynamic variables permits the decomposition of the state in such a way that only the part of interest need be dealt with.
If dynamic variables are integrated with the lexical environment, intractable dilemmas are encountered. (We have not considered here all possible such integration schemes, but the authors have found such difficulties with every such scheme they have examined.) We have therefore presented an interpreter in which environments for the two kinds of variable are separated.
