It is often considered a binary decision whether something is domain specific or not. Consequently, there are domain specific languages (DSL) and general purpose languages (GPL), there are domain specific and non-domain specific modeling tools, there are domain specific and non-domain specific methodologies etc. In this paper we argue, that domain specificity is not a hard decision, but rather one extreme on a continuum. We also argue that many systems can be more efficiently described with a mix of domain specific and non-domain specific abstractions. This view of the world has consequences for languages, tools and methodologies, specifically the ability to modularize and compose languages. Additionally we outline these consequences and provide an extensive example based on embedded systems.