Re: Rare Events One And All (@ tfb)
"their infrastructure is a chaotic mass of history and complexity which they can't just do a clean-sheet reinvention of because no-one really knows how a lot of it works except that it does, and it is just really, really hard to predict what is going to go wrong as a result."
Going back to your original well made point - the thing is, we see from thebigG, MS, AWS etc. that their infrastructure is also a mess of complexity. I blame (as I suspect you do with pointing to the number of Linux kernel lines) the amount of code being written in ever lower cost, lower experience shops.
I work in an industry riddled with ancient technology beating up against new tech. I work in old tech and I'm distinctly unimpressed with the new. The chaps programming it care mostly about writing pretty code; they don't care to understand the business, they don't care whether it works or not, just that it is leading edge. Performance is always a hardware problem to them. The languages themselves are opaque and insecure by design. Each environment seems to be hand crafted to be different to every other environment that's gone before. Communications between systems is a black hole fiasco (I suspect the Barclays issue was MQ related, I too have seen roque MQ messages block an entire network as they block every listener and no easy way to spot where the blockage is or where the bad packets are coming from).
And agile means there's no architecture, no low level technical direction. "As a user I want a banking mobile app that lets me check my account balance". Architecture seems to be limited to specifying components as if bricks, mortar, wood, nails are all all you need to design a house. "As a user I want a shelter that keeps me warm in the winter" - welcome to your brick oven; no windows...