The Knowledge Industry Survival Strategy Initiative (KISS)

The commercial benefits claimed for software based on Domain Specific Languages are well documented. Many DSL tools exist and are being used as point solutions. Tailoring of notations to the specific application domain and combined use of several languages define the nature of the approach, and constitute the source of the achievable benefits. Unfortunately data representations and the mechanisms used to integrate languages tend to be highly tool specific. This compromises the use of DSLs in building tool chains that may contain components from several suppliers. KISS is an industrial initiative that aims to define a set of fundamental principles for the design and integration of DSLs and to provide a framework for the development of DSL tools that achieves interoperability. 1 DSL: Benefits and Problems The main motivation for the use of a DSL [1] is the desire to express problems in a compact form that reflects the natural terminology of human domain experts, and that is easily accessible to software tools. In short, DSLs are raising the level of abstraction of software specifications and of knowledge representation in general [2]. When DSLs are used to formalize the results of domain analysis, the result is a clean separation of concerns in the problem space [3]. This is a major advance over aspect oriented programming, where separation of concerns is only achieved in the solution space. The value of a DSL increases with the intuitiveness of the concrete syntax. Visual and graphical elements may be needed to increase usability, and often such languages are referred to as domain specific modelling languages (DSML). The level of interoperability between current DSL tools is comparable to the level of interoperability between CASE tools in the 90s. To increase the popularity of DSL based approaches, this needs to change. With the extensive use of outsourcing and with the increasing investment in open-source software, software development has become highly decentralized, and an assumption that all parties in a global software supply chain will use identical tooling is simply not realistic. As a result today’s software supply chains are much less automated than supply chains in other, more mature industries. For example, the typical outsourcing relationship between two software manufacturers is based on informal specifications. Either the tooling used by the outsourcer is unknown to the customer, or if the customer is more discerning, the outsourcer sets up an environment that matches the one of the customer, usually at a substantial cost (licenses, staff training, etc.). The inefficiencies are such that it is not uncommon to set up mixed teams with staff from both organisations, and to rely heavily on extensive travel and face-to-face collaboration to achieve the desired outcome. The challenge lies in identifying the ingredients needed to create highly automated software supply chains that minimise the effort to integrate new suppliers of specialized software artefacts. Meeting this challenge requires a significant increase in the use of formal (yet highly compact) specifications, and advanced tooling for creating, managing, and exploiting formal specifications to the fullest. Ideally, supply chain participants would be able to exchange formal language definitions and models, and would rely on a shared implementation of basic services for managing such model based artefacts. A Knowledge Industry Survival Strategy (KISS) 2 customer may then provide an outsourcer with definitions of specification and implementation languages, and with appropriate instances of specification models. The outsourcer should be free to choose which tools are most appropriate for producing a solution that meets the specifications. The KISS initiative proposes a pragmatic bottom up approach to interoperability that is driven by an open community. Existing standards for modelling tool interoperability lack public reference implementations of test beds that can be used to establish compliance, and they are focused on the Unified Modelling Language rather than the requirements of heavily DSL based approaches. This paper is structured as follows: section 2 introduces a consistent terminology for describing DSL tool-chain components; section 3 reviews the current approaches and standards that influence the construction and interoperability of DSL applications; section 4 states the KISS values and aims; section 5 describes the issues that influence general tool interoperability and section 6 refines these to DSL tools; section 7 describes how meta-models influence tool interoperability; section 8 identifies a collection of KISS compliance levels that are proposed in order to achieve interoperability; section 9 describes how KISS is organized and presents the KISS roadmap; section 10 places KISS in context and reviews related work.