News for hackers. Stuff that matters.
Re:London Stock Exchange hauled offline after major data pro... by bmj on Friday February 25, 2011 @09:34AM (#2262271)
Does this same logic apply to, say, a control module in a space craft, or the software behind an MRI machine?
Re:London Stock Exchange hauled offline after major data pro... by SoftwareMaven on Friday February 25, 2011 @11:00AM (#2262552)
Any time you are dealing with a system with that many users, you are going to introduce a lot of chaos (in its mathematical sense). Chaos is nearly impossible to test for because there are just too many permutations.

You are (probably) right that any particular piece of a trading system would be easier to build and test than the examples you gave, but in aggregate, I think it would be much more difficult.

Re:London Stock Exchange hauled offline after major data pro... by spitfire on Friday February 25, 2011 @12:51PM (#2263046)
No it isn't. It's a matter of maturity and mindset. The avionics industry has it - they use Ada and formal methods, most of the rest of the software development world doesn't. and they pay the price. Honestly I questioned the competence of the LSE when they outsourced their development for the first windows based trading system to a consulting shop.

Simply put, they outsourced a core, critical part of the UK's financial infrastructure to boobs. However, I do see some hope in haskell and ocaml entering the financial market.

Re:London Stock Exchange hauled offline after major data pro... by nialo on Friday February 25, 2011 @10:20AM (#2262419)
Those are orders of magnitude easier to test.
Re:London Stock Exchange hauled offline after major data pro... by CWuestefeld on Friday February 25, 2011 @10:54AM (#2262528)
I'm not sure about that. The testing problems are different, but not harder.

An exchange has to deal with massive scale, but isn't hugely complex (relative to a spacecraft, which deals with fault tolerance and integrating many different inputs; or the mathematical complexity of medical imaging).

It seems to me that most aspects of scale aren't too difficult to emulate, given sufficient computing resources to generate the load.