2015-09-27

Test first, last, before, after, and in between ☺


Tom Breur
27 September 2015


I happen to like testing. Sure, it can be tedious. It can be frustrating. And often it may seem unnecessary. Especially when all goes well. For me, personally, somehow I get great satisfaction knowing I created something (or am about to deliver something) that makes me proud. Part of the enjoyment comes because I have verified that it works.

What is it about being in a hurry that seduces us (me, too, don’t get me wrong!) to skim on testing? How does time pressure (real or perceived) make us loose perspective on whether we should test something, or not? You have all heard the reasoning: “it’ll probably be allright”, “last time it worked”, “we probably won’t need that feature, anyway”, and the whopper: “It’s such a small change, what could possibly go wrong?”

After the “Flash crash” on May 6, 2010 at 2:45, it became clear that “innocent” stub quotes that were “just” in there, but weren’t going to be used anyway (!), had played a key role in spiraling the robots of High Frequency Traders out of control. These stub quotes had been set to arbitrarily low values (so as not to “interrupt” the software during testing, btw), because they weren’t needed anyway. Oops. That turned out to be an expensive error, $ 1 Trillion market value. That will remain a very difficult world record to beat…


No comments:

Post a Comment