Discussion notes on extending the McGPLE 08 paper into a journal submission

Focus on one case study: Three case studies are too much because there is not enough space to present an understandable description of the case studies. We focused on three case studies as a way to provide an informal proof that the research questions were actually representative. For the journal paper we can now refer to the McGPLE paper to say that we have investigated multiple case studies, but we will only present one case study as an illustration of the research questions. Focus on variability modeling: write a paper that focuses on research challenges on variability modeling and not that much on software architecture or implementation technology for self-adaptive product lines The paper should make also a tangible contribution besides the research agenda alone. In a sense, addressing research question 6 by making a classification or taxonomy of key concepts for self-adaptive product-lines seems really a good way to make such tangible contribution. Do we miss some research questions (wrt variability modeling?) If we look for example at the related work section of the paper, drew up Arnaud, we should have a look at these papers to see if these raise important challenges. Unclarity wrt to RQ4 (how to specify constraints in order to avoid underspecification): What is the definition of constraints? We used constraint as being a first-class representation of anything you would like to specify about context-dependent adaptations in a language; this involves both dependencies between contexts and adaptations, but also conflicts between adaptations, binding time conditions, relationships between contexts..So it is more a rule expressed in a formalism that is generic enough to capture the semantics of all these elements. Is it a problem of underspecification or overspecification or both? Unclarity with respect to RQ3 (How do NF concerns constrain the execution of context-dependent adaptations). Maybe it should be specified more towards specifying constraints on (re)binding (e.g. when is it safe to adapt)

Selection of a good candidate for the case study. We list the requirements for a good case study RQ1,5,6-inspired: It should contain a lot of context-dependent adaptation logic, this means: (i) a lot of contexts (variation in contextual information), (ii) a lot of dependencies between context and adaptations, (iii) the number of possible product configurations (combination of particular adaptations) must be representative , (iv) adaptations should be realizable as code-level variability at the implementation-level (not just by means of parameterization), (v) a broad range of different types of contexts and adaptations. RQ2-inspired: Ideally models of product-line variability and/or models of software variability already exist RQ3-inspired: The case study should involve constraints on when it is safe to rebind. (With rebind we currently mean replacing an already bound variant with another variant at run-time). RQ4-inspired: The case study should illustrate the problems with under-/overspecification that appear when specifying context-dependent adaptations with traditional formalisms for reasoning over feature configuration (e.g. first-order logic based formalisms)

Journal target: Possible alternatives Closest venue is workshop on Software Engineering for Self-Adaptive Systems (SEAMS). We should try to focus on journals where the workshop results are being published Special issue on Feature-oriented software development in journal Science of Computer Programming (submission deadline probably march 2010)

Action points Approve/Comment above directions for extending McGPLE paper (Arnaud) Propose candidates for case study (Eddy, Engineer, Arnaud) Propose candidates for journal target (Eddy, Engineer, Arnaud)

wp3/minutes.txt · Last modified: 2010/03/30 10:18 by eddy.truyen