On 1st November 2009, we released Deliverable 3.2, First version of the DSL based on the model developed in WP2, for review. You can find it here. This document outlines the process we undertook for examining current package meta-installers and exploring how a distribution agnostic, Domain Specific Language (DSL) could help in detecting certain types of failures not already identified by the current breed of package installers.

We firstly looked at the state of the art systems and re-visited the work carried out in previous Deliverables in the Work Package 2 and 3 streams, see the work streams. Work Package 2, primarily focuses on creating Models for the description of software artifacts and the upgrade-ability process, see D2.1. Work Package 3, which this document is a part of, addresses transactional roll-back and upgrades for Free and Open Source Systems (F/OSS), see D3.1.

Using the models, we aim to maintain a model-based description of the system and be able to simulate upgrades in advance on top of the model, to detect predictable upgrade failures and notify the user before the actual installation occurs and the system is affected. More generally, the models are expressive enough to isolate inconsistent configurations that are not currently expressible through the meta-data of inter-package relationships. With the definition of the first version of the language we aimed at creating a DSL that represents maintainer scripts and that predicts their effects on systems.

The language we used to represent the transformations between configuration states is that of the ATLAS Transformation Language (ATL), see the Eclipse documentation. We explain the motivation behind our selections within this deliverable. The MANCOOSI project, to be successful, requires feedback from industry, research partners and any interested parties.

If you have any feedback or comments please let us know as we are looking to improve upon the DSL for future versions.