Sure, that software's been tested

In the late '80s, I worked at the R&D facility for a PC software company. At that time, it was the largest PC database software maker and the third largest PC software company behind Microsoft and Lotus. Following the phenomenal success of their previous version, the company was working hard to put out the next version. Four years in the making, it was still undergoing serious feature creep as the marketing wonk

In the late '80s, I worked at the R&D facility for a PC software company. At that time, it was the largest PC database software maker and the third largest PC software company behind Microsoft and Lotus.

Following the phenomenal success of their previous version, the company was working hard to put out the next version. Four years in the making, it was still undergoing serious feature creep as the marketing wonks demanded ever more features to be grafted onto the basic framework. Since the CEO of the company's only previous experience was marketing soap (the kind that makes bubbles), he was ill-equipped to determine the quality of the product. In fact, rumor had it that he couldn't even start it up. Despite this, he was among the highest-paid CEOs in the country.

As a result, company ended up shipping the product while it was still full of bugs, which effectively destroyed the company when the product became the laughingstock of the industry.

The whole story is a cautionary tale about software development and the pitfalls in it. Two senior developers took to handing the CEO a copy of The Mythical Man-Month each time they saw him. He would thank them and leave the book on the table as he left the meeting. If only he'd taken the time to read it…

But that's not my story. This anecdote is about one of the testers for the product.

Each area of the product was rigorously tested using (for the time) an extremely sophisticated automated system that would run a set of test scripts on a test machine, then store the results from the test on a central server. Anything that produced an error would generate a work order for the coders to fix. The scripts were all written in the product's scripting language.

One particular tester was hired as a contractor to write test scripts to wring out all the math functions. More than a hundred scripts of his were run on each night's new build, and while other areas would have various problems, everyone was pleased that the math functions seemed really solid -- all the tests showed there were no math errors.

After a while, some of the senior developers got a little suspicious of this rosy picture and went in after-hours to look at the tester's scripts. To their amazement, each and every one of them consisted of a single line "Return True" (or the actual code equivalent). This tester had tested nothing, but instead just reported everything was perfect.

He was fired the next day. I think he became a dry cleaner.

Personally, I think it should have been a red flag when the guy went out and purchased a matching pair of "how to code" instruction books for version 2 of the product and set them side by side on his shelf. I guess he thought that, put together, they'd teach him all he wanted to know about version 4!

Never underestimate how little work someone will do if they can get away with it.