I'm not the guy who has created this statement but I love it, and I want to tell a little story about how really correct it is.
Some time ago I've worked on a project where we had to integrate a very old procedural language's creations into a Java application. I mean, old must not mean that its overhauled, but in this case it absolutely was.
In order to integrate or refactor whatever you need you always should create test cases and scenarios to check if your result works as expected or if your refactoring result does the same thing the smelling one did before. Additionally, you should measure a couple of things concerning the static code quality of what you do such as cycles, dependencies etc. This is the rule and the bible, so I don't want to go deeper with the reasons for it here.
We suggested to the customer who was technically responsible for all that old stuff to wrap his stuff around with tests an to (re)write a couple of standard metric measurement mechanisms like dependency measurement to check out where we are and how good or bad it is in order not to do the same mistakes later integrating our stuff.
And the answer was: stop doing that! We know how good our software is, and we do very much for it to get better every day. I asked him how he does this if he even cannot measure a simple dependency cycle in his stuff there, and after I did measure I've found roundabout two hundred of them in there. The aswer was they do it pragmatically without those scientifical methods.
Well, may be it's ok, but it smells and works only in the current constellation somehow. The source code degenerates and erodes continuously, and the only reason why it didn't already explode was the very small feature capability of this old language. Controversial? Not a bit, but somehow logical.
The fact is that you don't know when it crashes if you don't measure its static quality. And you cannot say you improve its quality working only with the gut instinct - you play the Russian roulette.
Measure the quality of you code, however old its base technology is! You cannot control what you cannot measure!
