The 10th International Software Product Lines Conference (SPLC 2006) is now less than three weeks away and I’ve been putting the finishing touches to the tutorial I’m running at the conference on Tuesday 22nd August - The Scoping Game. The slides went off to be printed last week, flights and hotels are in place so, barring last minute hitches, everything is now in place for the trip.

This has given me time to think about what other sessions I might attend…

I have no definite plans for Day 1 yet but the tutorials on Product Line Requirements Engineering and Creating Resuable Test assets in a Product Line should be useful. The workshop on Agile Product Line Engineering could also be worthwhile.

Day 2 begins with my tutorial session and I’ll probably take the afternoon off, although I am tempted by Don Batory’s session on Feature Modularity in Product Lines.

Keynotes from Gregor Kiczales and Carliss Baldwin begin Days 3 and 4 of the conference. I’m hoping that Gregor’s session will tie into some of the Early Aspects work I talked about at SPA 2006. Carliss’ talk focusses on Product Line economics - an essential issue underpinning the whole business case for Product Lines - which relates to the material in my Scoping session.

In terms of other sessions I find that I get the most value from Panel sessions and Experience Reports. It’s generally in these types of sessions that concrete advice on Software Product Line adoption that I can use with clients emerges. This isn’t to knock any of the research that’ll be reported in other sessions - it’s more a matter of using my time to best effect as I can usually read up on the research reports in the conference proceeedings and follow-up with the researcher later if necessary.

One disappointment this year is that I have to miss the (usually entertaining) Hall of Fame session as I have to travel to the West Coast in the afternoon of Day 4 for work.

If you haven’t registered yet there are places available (Early bird registration closes on 9th August though) so I hope to see some of you there. Come and say hello.

The July / August 2006 issue of Embedded Systems Engineering magazine has a lot of coverage on Modelling and Code Generation for Embedded Systems. In the editor’s words “Code Generation may not produce the quality of code that can be written by the very best programmer, but the best tools produce competent code that does the job, and they produce it quickly.” Six articles, primarily written by vendor representatives, look at different aspects of the topic:

“SysML hits the home straight” gives a short overview of the SysML modelling language - an extended subset of UML that includes special support for requirements traceability and Blocks - essentially a general purpose way of modelling system hierarchies. Worth a read if you just want a flavour of what SysML is and if you’re convinced that Visual modelling languages are the way to go.

“Modelling in Eclipse” is promisingly titled and begins with a brief description of the Eclipse development framework but the focus of the article is on Executable and Translatable UML (xtUML) a subset of UML with a defined execution semantics, that can be translated into an implemetation, rather than a general survey of the various modelling options available for Eclipse.

“Generating full code with DSM” makes a good argument for Domain-Specific Modelling languages to support code generation. The idea behind domain-specific languages is to allow you, the developer, to control what code gets generated when and to do so by allowing you to create a little (or big in some cases) language that is best suited to capturing your design intent. This approach should give you more confidence in the generated code than that generated using a black-box generator approach.

“Developing platform-independent embedded applications” makes a case for Platform Independent Modelling as the basis for Code Generation. This is the approach advocated by the OMG’s MDA initiative which I’ve talked about before.

“Systems & software modelling for safety critical environments” argues that modelling is attractive in settings where time-to-market is critical and development cycles are short but notes problems of determination and ambiguity in model-based approaches. SCADE, a domain-specific language with roots in the aeronautics industry, is introduced as a tool which has been built with these problems in mind.

The final article, “In-depth: System modelling and verification“, looks at these important topics and how they are being addressed by tool vendors.

All in all an interesting issue.