What do we want from a high-level language?

In this paper, we discuss some of the issues involved in choosing a better language in which to write computer algebra systems. We conclude that no wide-spread existing language is superior to those we currently use, and mention some of the alternatives.