Language workbenches (LWBs) are tools to define software languages together with tailored Integrated Development Environments for them. A comprehensive review of language workbenches by Erdweg et al. (Comput. Lang. Syst. Struct. 44, 2015) presented a feature model of functionality of LWBs from the point of view of the ``languages that can be defined with a language workbench, and not the definition mechanism of the language workbench itself''. In this vision paper, we discuss possible functionality of LWBs with regard to the language definition process. We have identified five groups of such functionality, related to: \emph{meta-definitions}, \emph{meta-modifications}, \emph{meta-process}, a \emph{language workbench itself}, and \emph{programs} written in languages defined in a workbench. We give an example of how one of the suggested features—\emph{specification of explicit dependencies between language concerns}—can be designed based on our vision.