Including diagnostic information in configuration models

Abstract

This work presents a new formal model for software configuration. The configuration knowledge is stored in a configuration model that is specified using a rule-based language. The language has a complete declarative semantics analogous to the stable model semantics for normal logic programs. In addition, a new method to add diagnostic information in configuration models is presented. The main idea is to divide the configuration process into two stages. At the first stage the user requirements are processed to check whether there exist any suitable configurations in the configuration model. In the second stage unsatisfi- able requirements are diagnosed using a diagnostic model. The diagnostic model is constructed from the configuration model by adding a new set of atoms that represent the possible error conditions. The diagnostic output also explains why each problematic component was included in the configuration. As an example, a subset of the configuration problem for the Debian GNU/Linux system is formalized using the new rule-based language. Both configuration and diagnostic models of the problem are presented. The rule-language is implemented using an existing implementation of the stable model semantics, the Smodels system.