Wednesday, September 25, 2002
I picked up a link from Jiri's web site to JoelonSoftware an article noting that NASA is cancelling a project to build a new launch control system that has already been running for 5 years and has cost $273 million. I guess NASA is, strictly, speaking public sector - but not that long ago they were held up as the benchmark for "good" software development (SEI level 5 and all that - I rate this equivalent to ISO 9001, or the concrete life jackets standard - i.e. it doesn't have to work as long as you build it consistently over and over and over again). Joel's wonders why are we still building [multi] billion dollar monolithic systems ... and why are they failing? I've been wondering myself about this kind of thing for a while - across every public sector organisation in the world you will find piles of billion dollar/pound/euro projects building mammoth new IT systems to deliver increasingly finer-grained taxation systems, benefit delivery mechanisms and so on. These systems have web front ends that contain rules engines and often are cloned by third parties (like Sage, Oracle, Peoplesoft) who need to deliver similar functionality to allow companies to prepare their accounts. So, what we really have is a front end (with some rules) ... a back end (with the same rules) and a general ledger (that makes/receives payments and keeps the books). What I don't follow is why we have n front ends (where n is very large), m back ends (where m is probably 1/3 n) and m general ledgers ... what we need is m, 1, 1. Or, at least one general ledger and one generic rules system that exists in "space" (i.e. out there, not in here) and that can be called upon by third parties and government systems alike. This needs some rethinking about where the data is, how it's stored and who looks after it (dare I say that the citizen could be trusted to look after their own data and make sure it's up to date?). I wonder if anyone is doing this? Anyone know?
Posted by Alan at Wednesday, September 25, 2002