I ran a variant of my talk When Good Architecture Went Bad at the Embedded Masterclass recently. This is a free, one day event with technical talks and a small exhibition aimed at embedded engineers.

This year there were actually two Masterclasses, one in London and one in Bristol and the combined attendance was around 140 - despite problems on the London railways conspiring to keep a few people away.

A couple of questions that came up during my sessions:

• Do real-time software systems suffer from architectural decay less than other types of software? - I have a gut feel that no, they decayed just like other software systems despite more rigourous verification methods being used in many cases. This is because architectural decay comes from change - change in functionality, change in staffing, change in tools and technologies and I don’t feel that in general real-time systems are less prone to these types of changes.
• Does UML help? - Not being a big fan of UML I passed this one onto Kevlin Henney who happened to be attending the Bristol Masterclass. His view - yes and no. Yes if you use UML to document what you have implemented (and so communicate the architecture), no if you rely on a UML document to act as a guide for implementation. (apologies to Kevlin if I misremembered this)

Being in Bristol also gave me the chance to meet up with some of the local ACCU members. We went to the Old India restaurant in the city’s former stock exchange - great food in a great setting with great company. Thanks to Tony Barrett-Powell for arranging this at short notice.