On Sun, Sep 28, 2014 at 12:22 PM, Greg Lindahl lindahl@pbm.com wrote:
A third source is companies with homegrown code deployed on CentOS servers and poor-quality test suites. They tend to be in the "omg never change anything unless forced at gunpoint!" camp. It's an unfortunate situation, and it can cost a lot of money and time to fix.
Or even with decent test suites you recognize that you can't perfectly emulate a live internet-connected production environment. Or you've been burned by updates that did break things and the time it took to find a workaround. There's just no getting around complicated systems being complicated.