Yes – I can appreciate your comments. I also agree moreover that we “went out too long” with this release. While I have been successful with implementing scrum in several teams in my current position, the higher up you go in the organization, the harder it is to convince people to change. I actually had a number of conversations that are not to dissimilar from your comments.

The salient bits to me are:

- humility is important.
- unit tests are imperative, but will not prevent all bugs.
-your velocity can be killed by the ‘outlier’ event

As your comments help illuminate; N smaller projects running under a non-agile umbrella – is not agile.

I am making headway still – and I think I have managed to compel people throught this particular project that you MUST be more iterative, must be getting end user feedback earlier and must simply disperse the risks of project delivery into smaller bites.

All great food for thought – thanks again for the comments.

-Chris

No – the bug was not concurrency related. It was related to a deep clone implementation. Essentially, we have a layer that does deepcloning of some object graphs. Part of the graph was pointing, to something else’s internals. It was a very C++ throwback set of circumstances.

We actually have lots of tests for this particular app that deal with its multi-threaded aspects. These sort of fall into the ‘integration tests in a unit test harness family’ moreover, but they were helpful and I believe effective. I would assert that writing good concurrent software is hard, but doing so with tests certainly makes it all a lot easier.

-Chris