An alternative approach to I/O

We propose a form of input and output for functional languages that is in a sense \emph{orthogonal} to the actual computation: certain input and output directives can be added to a completed, fully working program text, and they do neither disturb referential transparency nor necessitate to change the types of the program text. The input and output directives change the order of evaluation as little as possible (lazy evaluation remains lazy), though there is sufficient control over the order in which the input and output actions occur to make it acceptable for the user. The basic idea is that a value which is written out explicitly in the program text by way of typical example, is replaced by a special constant that asks the user to type in parts of the value, as needed by the computation. The mechanism seems suitable for a large class of so-called interactive programs.