I saw a zdnet report last week quoting Reuters about a $105 million dollar computer project write off at the FBI. Here's the key bit:
The FBI has designed a new computer system to replace a failed $170 million one aimed at helping agents share information but it will not be ready for use until the end of 2006, the FBI director said Tuesday.
You know what the most critical error was in the one they wrote off? the Inspector General's Audit report cites all the usual suspects, but missed the fundamental problem completely. What killed that project's hope of success right from the beginning was the simple fact that they hired the people and bought the hardware before figuring out what the software was supposed to do.
You know how well they're going to do this time? You guessed it, they've got the architecture before the software again.
You'd think organizations would learn, but they don't -and the budgetting process doesn't help the people who do. Basically it's almost always easier to get hardware approvals through for the fiscal year right after disasters, startups, or major re-organizations, so a lot of these guys end up with hardware budgets they have to commit long before they know anything about the software they're going to need.
What you get out of that is exactly what happened -and will happen again- at the FBI: since they don't have the software figured out they assume Wintel is the answer, get the people, hardware, and operating system licences to go with that, and then discover they're on a rockslide that can only go one way - downhill.
Want some free advice Mr. Mueller? Get past your preconceptions. Most of your problems are coming from your need to blend real time integration with security in an environment with 30,000 or so Wintel PCs. Learn something from your failures: TV and advertising to the contrary, it can't be done without making the system useless. So go ask Bill Voss for a tour of Sun's national data center and have him show you how putting a Sun Ray on every desk could give you a successful project in as little as six to eight months -complete with direct Linux and Solaris laptop integration and a price tag that's likely to be 60% less than what you hope this round with the client-server money machine is going to cost the taxpayer. And, no it's not the bucks that are important here, what's most valuable to you is that the security and the real-time integration your people can't deliver on the PC are automatic, while most of the software you've been trying to invent is already there.