The programming and modelling languages currently used in software engineering generally have plenty of tool support. But although their syntax is specified using formal grammars or meta-models, complete formal semantic specifications are seldom provided.

The difficulty of reuse of parts of semantic specifications, and of co-evolution of such specifications with languages, are significant drawbacks for practical use of formal semantics. I have collaborated in the development of several meta-languages for semantic specification, aiming to eliminate such drawbacks: action semantics, and modular variants of structural operational semantics (MSOS, I-MSOS); this led to the PLanCompS project and to CBS, a meta-language for component-based semantics.

The components of language specifications in CBS correspond to so-called fundamental programming constructs (funcons). The main feature of CBS is that each funcon is defined once and for all: the addition of new funcons does not require any changes to previous definitions, and behavioural laws are preserved. In contrast to software packages, the definition of each funcon has to remain fixed after its publication.

As well as explaining how component-based semantics achieves these desirable pragmatic properties, and comparing its features with those of some other meta-languages, I will demonstrate the current tool support for CBS, which is implemented in Spoofax.

Peter Mosses is professor emeritus at Swansea University. He is currently visiting the Programming Languages Group at TU Delft.

His research in semantics stretches back to Strachey’s Programming Research Group at Oxford in the early 1970s, where he contributed to the development of denotational semantics, and implemented SIS, a system for running programs based on their semantics. He was based at Aarhus University, Denmark, from 1976 to 2004.

The main focus of his research has been on pragmatic aspects of formal specifications – especially modularity. This led to the development of action semantics, MSOS (a modular variant of structural operational semantics) and component-based semantics. He is a principal investigator in the PLanCompS project (Programming Language Components and Specifications), He was also the initial coordinator of CoFI, the Common Framework Initiative, which designed the algebraic specification language CASL.