User Story Based Engineering
In other blog posts I’ve talked about using Multi-stage Continuous Integration to scale Continuous Integration for use in multi-team environments and using streams or branches to model your workflow directly in your SCM tool. While both of these approaches work well and provide a lot of value, they can also present a new challenge: how to associate enhancement requests and defects, which can be lumped together using the term “work items,” with the changes that implement them?
If you are doing mainline development, life is pretty simple. On check-in enter the enhancement request ID or defect ID in the comments. Once that association is made, it is fairly straightforward to figure out which enhancements and defects have been checked in. On the other hand, the more people you have checking in to the same branch, the less likely it is to be stable, even if you are doing Continuous Integration. So, how can we maximize the benefits of using streams to represent process and integration stages and still have the ability to match work items to the work done to implement them?
Representing Process and Integration Stages with Change Packages
(Let me first point out that while I’m using the generic term “work item” here, in an Agile team one would most likely use the term “User Story.” Both apply equally well in this context, but I’ll stick with work item for now.)
There is an SCM concept that maps to a work item which can be tracked from place to place, and that is a “change package.” It is called a change package because it represents something which, once defined, can be applied to multiple places and tracked regardless of how many different places it has been propagated.
Change packages can be promoted individually or in groups from stream to stream. That way, as work is promoted from stream to stream you have a record of the content of each stream from a high level perspective instead of a files perspective. A list of files is not very meaningful or useful. On the other hand, change packages allow you to ask questions like “which user stories are in the stream that represents all user stories that are ‘done’,” without having to resort to mainline development.
