D2's post on statis (it's the new change) and crap government IT brought something to mind. Dan mentions the success of the Bank of England-run Crest settlement system for the London Stock Exchange, contrasted with the hellbroth of disaster the NHS National Programme for IT is descending into. One thing I think he should have mentioned, but didn't, is the role of institutional memory.
He correctly points out that managerial stupidity loves the idea of change and the notion that past history is no guide. Very true. But one of the worst things about this is that the accumulated knowledge in an organisation is also irrelevant. In his example, the Bank profited from its experience - you didn't see any Big Consultants in that post - and succeeded.
On the NHS IT project, the failures so far are iSoft, which illustrates the fallacy of thinking that dynamic young start-ups necessarily know anything, and Accenture, which illustrates the fallacy of thinking that management consultants know anything. The only sections successfully delivered are those being built by BT, which has been doing big networks and big databases for donkeys' years and working with the public sector for as long. It's possible BT's job was easier. They did the so-called national spine (some fine distinction between a spine and a backbone network..), which is essentially a big VPN deployment over their MPLS network and some data centres. So far, so cookbook engineering. But BT also has one of the much more complicated regional integrator contracts, and none other than the London Region one. That hasn't gone to ratshit yet, as far as I know, so they must be doing something right.
Similarly, the disaster that was Railtrack had a lot to do with listening to people other than the people who knew what they were talking about. The BR engineering department was asked to piss off out of the West Coast Main Line project so consultants could prepare cost estimates more congenial to the government, and then the consultants were asked to work out how to run the company (bang goes the BR operations department). They turned out to be as wrong as they could possibly be.
IBM turned up the old documentation from the days of modular mainframes when they designed the first Bladeservers. Some engineers on the project remembered them.