Thursday, August 30, 2012
By Anthony F. Voellm (aka Tony the perfguy)
It’s amazing what has happened in the field of test in the last 20 years... a lot of “art” has turned into “science”. Computer scientists, engineers, and many other disciplines have worked on provable systems and calculus, pioneered model based testing, invented security fuzz testing, and even settled on a common pattern for unit tests called xunit. The xunit pattern shows up in open source software like JUnit as well as in Microsoft development test tools.
With all this innovation in test, there’s no wonder test is dead. The situation is no different from the late 1800’s when patents were declared dead. Everything had been invented. So now that everything in test has been invented, it’s dead.
Well... if you believe everything in test has been invented then please stop reading now :)
As an aside: “Test is dead” was a keynote at the Google Test Automation Conference (GTAC) in 2011. You can watch that talk and many other GTAC test talks on YouTube, and I definitely recommend you check them out here. Talks span a wide range of topics ranging from GUI Automation to Cloud.
What really excites me these days is that we have closed a chapter on test. A lot of the foundation of writing and testing great software has been laid (examples at the beginning of the post, tools like Webdriver for UI, FIO for storage, and much more), which I think of as Testing 1.0. We all use Testing 1.0 day in and day out. In fact at Google, most of the developers (called Software Engineers or SWEs) do the basic Testing 1.0 work and we have a high bar on quality. Knuth once said "Be careful about using the following code -- I've only proven that it works, I haven't tested it."
This brings us to the current chapter in test which I call Testing 1.5. This chapter is being written by computer scientists, applied scientists, engineers, developers, statisticians, and many other disciplines. These people come together in the Software Engineer in Test (SET) and Test Engineer (TE) roles at Google. SET/TEs focus on; developing software faster, building it better the first time, testing it in depth, releasing it quicker, and making sure it works in all environments. We often put deep test focus on Security, Reliability and Performance. I sometimes think of the SET/TE’s as risk assessors whose role is to figure out the probability of finding a bug, and then working to reduce that probability. Super interesting computer science problems where we take a solid engineering approach, rather than a process oriented / manual / people intensive based approach. We always look to scale with machines wherever possible.
While Testing 1.0 is done and 1.5 is alive and well, it’s Testing 2.0 that gets me up early in the morning to start my day. Imagine if we could reinvent how we use and think about tests. What if we could automate the complex decisions on good and bad quality that humans are still so good at today? What would it look like if we had a system collecting all the “quality signals” (think: tests, production information, developer behavior, …) and could predict how good the code is today, and what it most likely will be tomorrow? That would be so awesome...
Google is working on Testing 2.0 and we’ll continue to contribute to Testing 1.0 and 1.5. Nothing is static... keep up or miss an amazing ride.
Special thanks to Chris, Simon, Anthony, Matt, Asim, Ari, Baran, Jim, Chaitali, Rob, Emily, Kristen, Annie, and many others for providing input and suggestions for this post.
Labels: Tony Voellm