As I’ve noted previously the University of Oxford runs a specialist course on Software Product Lines as part of its post-graduate Software Engineering programme.

Roberto Lopez-Herrejon was kind enough to invite me to give a talk about pure::variants to the class last week and I was very happy to oblige.

It was quite a different experience for me speaking in front of an audience that is drawn from industry but has an academic setting.

The questions I got reflected the nature of this audience:

• Does the Feature modeller handle cardinality? - yes e.g. selecting n from m in an OR relationship, feature attributes may also take on a value from a range of values
• What’s the difference between a relation and a restriction? -  Semantically they are not that different. Relations like conflicts /requires etc. can also be expressed with restrictions / constraints. However, relations have a limited expressive power, therefore when there are problems they are much easier to analyze. Also, due to the way pure::variants handles target selection for relations mistakes are much less likely. Relations can also have other (not configuration related) semantics.
• Does pure::variants validate Feature Models (To ensure at least variant can be produced?) - The extended autoresolver knows how to check this property.
• Why use PROLOG as the rules language and constraint checker? - available , high-performance tools, good at this type of problem, syntax not too difficult (rule definitions)

After the talk I was thanked and given a University of Oxford T-shirt. Not sure I can wear this around town (Cambridge) though.

Roberto has run the course a couple of times now and plans to do so again in the future. Drop him a line if you’d like to participate - although you probably have to register for the MSc.. I hope to speak there again - if invited - and we should be able to run a pure::variants tutorial there next time as opposed to just a talk/demo.