Selenium Conference India 2018 Experience

Early in February 2018 I saw a tweet from Manoj that Selenium Conference was coming back to Bangalore. I booked the conference ticket and the flight within a few minutes. This note is the two conference days. There is no separate write-up of my own session. It lives here.

About SeConf

SeConf is a community event, run twice a year by volunteers. The shape is three days: one hands-on workshop and two days of conference. It is not a Selenium product show. Most of the conversation is about testing, and some of it is about when not to use Selenium.

Day 1

The opening keynote was the yearly “selenium state of the union”, this time Simon Stewart and Manoj Kumar in a conversation: how the project started, how it moved from 2007, and what else in the industry moved with it. Manoj made the open-source point directly. The project needs people, and contribution is code, documentation, or helping someone on a forum or a blog.

Raj Subramanian spoke on “How AI is transforming software testing”. The example I wanted to try was dynamic locators instead of one static locator, so a UI check can keep running and report a lower confidence instead of dying on the first markup change.

Jatin gave a short talk on continuous test data generators: volume, diversity, and why a UI suite needs both. He also said, in the hall afterwards, that the approach had found bugs and widened coverage.

Gaurav Tiwari spoke on “Including voice command related tests to your automation Suite of Selenium/Appium test”. The room was full, well over a hundred people, because voice was new to most of us. He described the techniques he used at Amazon, and an open-source project on GitHub so the idea was not only slides. I like a talk that leaves something you can run.

The next talk was mine: Testing as a Chat. I talked about extending ChatOps into testing, driving different kinds of tests through chatbots, and what that does for collaboration. A few examples were on GitHub. The questions afterwards were specific, and the room was curious rather than polite. The original note said a separate post and a user guide would follow. They did not. This section is the record of the talk.

Dan spoke on six years of building and maintaining Appium: a lightning talk that became the usual way teams automate mobile. The lessons were about growing an open-source project, not about a feature.

The evening keynote was Ameen Haque, from StoryWallahs, on “The Art Of Business Storytelling”: people move on stories, and the craft is useful in the rest of the work, not only on a stage. It was the most fun session of the day, and a few of the lessons stayed.

The VU team closed the day with games that were louder than the description suggests. It was also Titus’s birthday. He maintains the Selenium Ruby bindings, and the room sang. Dinner was the other half of the networking.

Day 2

I changed hotels and underestimated Bangalore traffic, and I missed Maaret’s keynote on exploratory testing and the speaker photograph. I had already heard enough about the keynote from other people that the video was the next stop.

Jim Holmes spoke on “Changing Testing Culture in a Ginormous company”: making a change that sticks in a place where culture moves slowly. Those stories are the ones I look for. A cultural change is expensive and, when it works, it is the thing that lasts.

Marcus spoke on microservices. I had seen that shift in my last two companies, and I still left with a few points. The talk was about testing, not about Selenium.

Dmitry spoke on “Mobile Visual Testing”, with enough live examples that the useful frame was the one he gave near the end: a visual test automation pyramid, and the order in which to adopt the layers.

The Q&A with the Selenium panel was the liveliest session. The questions ran from the project to the W3C spec, the conference, and who was on the panel. Same as day one: the last session did not empty the room.

Who was there

About 450 people, a bigger and more mixed room than SeConf 2016. Most were senior software engineers or QA engineers, many with eight years or more, and they were there to compare notes.

I saw people from the previous conference, colleagues from an earlier company, and a lot of new people. The speakers were easy to talk to. About a quarter of them were women, which was good to see in the lineup and not only in the audience.

Volunteers ran the event. Naresh Jain and the ConfEngine team shipped the app that carried the schedule and the introductions.

Bhuvan, from my own company and team, was there too. I was glad not to be the only one from the office.

Takeaways

• Visual checks had become normal work, and the argument in the room was that people are a poor oracle for a complex screen. A computer, used carefully, is better at seeing a visual defect.
• What you test should be chosen from how the product is used and from the history of your own results, not from a static case list alone.
• Machine learning and pattern analysis were starting to show up inside testing tools, towards automation that can judge as well as click.
• A functional result does not have to stay pass or fail. Raj’s dynamic locators were the example: a result with a confidence, not only a colour.
• Test infrastructure was the scaling problem. A few conversations were about AWS Lambda and Google Cloud Functions as a way to get more parallelism.
• Mobile, for a lot of the room, was no longer the unsolved platform. The interesting targets were TV, wearables, voice, chat, VR, and IoT, and Appium was already being stretched toward them.