We wanted to set up a small team who’d look after the areas that overlapped across projects, the common code, framework, that kind of thing. So we started experimenting with using index cards - we just had little summaries on them and we tried to keep them as small as possible, but they were still very big for what I’d call a task now.

One thing I noticed quite early on was that we had these project plans which were out of date almost as soon as we started and got worse as time went on. We had a massive requirements document which was out of date. There were also test plans which actually got written after the fact; they weren’t supposed to be, but most people did that because it made logical sense to them. The other thing I noticed was that the date and the time from when a project was first claimed to be complete to when it actually was actually printed onto a CD varied from six weeks to, in one case, 18 months.

How Agile works depends on the organisation. At times I’ve thought “Oh well, we were doing things wrong” but other times I’ve said “Well, we were doing things in a way that just isn’t right now.” It was right for the organisation then but it wasn’t right for the organisation now. We’ve evolved into a different organisation from the one that we were and therefore the way that we apply XP is different.

One of the things that we battled with through the team was when we brought in a dedicated test guy to work on our QA. I don’t think any of the other developers actually had any prior experience of working with dedicated test people. We had limited experience. But also, because we had this emphasis, we wanted developers to take ownership of what they were delivering. So we found that we never really evolved a very good way of working with this guy. He understandably eventually got fed up and he left. We decided not to replace him.

While [T] and I were working together there was quite a lot of freedom. We were, certainly for a period, quite freely refactoring things. When we recruited people we were very much focused on building a real live service. Suddenly things became a lot less fluid and refactoring became quite an issue.

I remember being challenged by the developers who were involved in that early stage. I know one of their contentions was that we were not refactoring. It became very clear that I didn’t get refactoring at that time. Some of it I was doing… but I didn’t get it enough, and I was definitely one of the obstacles to them being able to do some of the refactoring that they wanted to do.