15 September 2010

Software product line engineering

Software product line engineering is apparently about modelling variants and achieving formalism in feature descriptions. This is the conclusion I make after attending some workshops and the first morning sessions at the 14th International Software Product Line Conference in Jeju, Korea.
I presented an interesting  paper together with Håkan from Scania of how architects work with maintaining and updating existing architectures over time in the automotive industry. And we did not get a single question!
Last year when I presented another case study everybody was interested in the case and wanted to hear more from industry, but this time it didn't seem to interest the audience.
I find it quite difficult to find venues to present research based on industrial experience and not theoretical examples.

Besides that, I find the notion of feature used in many presentations different from what I am used to. To me a feature is something which is discernible for the end user, same as the definition found in the original work here. For example an adaptive cruise control is a marketable feature in a vehicle. But if I would model that similar to feature modelling prevailing here the model would consist of an optional radar, a compulsory engine, compulsory brake, etc.This means a much higher degree of knowledge about the realisation of features.


Andreas Graf said...

There is a number of automotive / embedded conferences here in Germany that might be of interest to you then - they value talks based on industrial experience, especially from end users.

The notion of feature might be different, but I think the example of prevailing model is representing the needs of the industry. Detailing the combinations of ECUs/functions can be done by feature diagrams (even it might not be a feature). See e.g. the work on CVL by MoSis.

Ulrik said...

I know our research is of interest of practitioners, at least in Sweden. Håkan and I have given 3 presentations internally at companies, and I'm working with setting up more workshops with local companies in Göteborg.
What I am concerned about is it seems the majority of the research community seems to be rather uninterested is certain areas which seem to generate a lot of interest of practitioners. This is the second conference I've been to where people from industry openly voice a concern that academic research is diverging from industry needs and not converging.

When it comes to what a feature is it might part be language problem of the entities we talk internally about in development at Volvo when it comes to standard, option or market dependent. However if I would open a sales brochure It is quite clear that automotive companies market an adaptive cruise control as a feature the end user and not a radar...
Thanks for the tip on CVL, I will read more about that even though it is not related to the core of my research