Wednesday, September 17, 2003
A lasting legacy?
I've been worried for a while now that all of our efforts in the world of e-government (and here I mean pretty much everywhere), where we have strived to find ways to get around or outright remove legacy systems (the ones that work, or the ones that have been there 20 years - whichever your view) might just result in us creating another legacy problem. The kind of things that are explicit in an enterprise architecture - a truly componentised, citizen-led, product as servant information technology base that is fully aligned to the (revamped and reworked) business processes - are not yet prevalent as I look around at who is doing what. Sure, there are pockets of creativity and sometimes whole swathes where people are doing the right thing. But, in big picture terms, the legacy we look like leaving is not much different from the one our predecessors struggled to keep going - albeit this one is doubtless in C++/Java/VisualStudio and what not instead of Cobol. We can't afford for our legacy to be "another legacy" so what do we do about it? How do we find the buttons to push that make the decision makers eyes' light up as they see what could be achieved with the right technology? How do we gethe business thinkers to realign their processes so that we can create some common components that support more than one business unit? How do we do that in an environment where, rightly, the money is tight and the bankers want to be sure that what we spend now is justified and will deliver real saves? The promise of online government is great, but the experience so far has been lacklustre. This is the main thread of the work I'm doing on my EntArch - with a direct focus on what do we have to do this time that's different from before so that we don't fall into the same trap. I'm pretty keen on a legacy that fully supports "unplug and replace v1.0" allowing all of the components or even sub-components (fragments?) to be ripped out, upgraded and replaced when new things become available, without having to poke into the monolithic remains of prior applications.
Posted by Alan at Wednesday, September 17, 2003