Abstract

A language tailored to the problem domain can focus on its idioms and jargon, avoiding clumsy, overly general constructs needed to support general-purpose language. The leverage provided by DSLs over conventional programming languages is often extreme; application engineers may specify as little as 2% of the code that one would need to program the same thing in a conventional programming language! But commitment to a DSL approach can be rather expensive. It is often difficult to know when to invest in exactly how much infrastructure support for a product or product family. All of the concerns that are germane to general-purpose programming language design and support may become important in the support of a specific DSL. At the same time, there is a wide spectrum of approaches to providing DSL support. This paper relates the various DSL design approaches to alternatives for tool support, providing a kind of DSL tool support selection framework, indicating where one might expect to need to invest heavily to obtain adequate support and illustrating the spectrum of tradeoffs and situations in which each is appropriate.