July 2015 Summaries

In this guest post by Ron Blanford, Yammer Product Manager, he discusses how the company uses hypothesis-driven development to improve their products. The team recently overhauled their iPhone publisher to make it easier for users to post photos on Yammer. Despite initial disappointing results, they decided to ship the feature as they believed they had created a better experience and refactored old code. This case highlights the importance of being data-informed but not completely reliant on data when making product decisions.

The text discusses the use of lock-based concurrency in Go programming, specifically for lazy initialization problems where a resource is expensive to construct but read often and written only once. It presents two approaches using read/write locks and sync.atomic package respectively, and finally introduces the sync.Once utility which encapsulates all the locking logic. The author emphasizes that channels should be the first choice for structuring concurrent programs in Go, and these low-level synchronization primitives are last-resort options.

In 2014, John and the author started working full-time on LaunchDarkly, a platform for feature flags, dark launches, canary releases, and continuous deployment. The idea was inspired by how companies like Facebook, LinkedIn, Etsy, and Netflix were rapidly iterating their software and releasing multiple times per day or hour directly to users. Feature flags and dark launching have been adopted by various large companies such as Flickr, Facebook, and Instagram. However, not all businesses could afford the infrastructure for these practices. LaunchDarkly aimed to provide this capability as a service at an affordable cost. Since its inception, the platform has gained customers, added engineers, and hosted its first Dark Launch meetup. The focus now is on continuous improvement and customer feedback.

The term "beta" has evolved from its original purpose of gathering feedback on product features to now signifying that a product is not fully ready for release. Companies like Microsoft, Instagram, and Google have adopted canary releases as an alternative approach to testing new features with a subset of users before full-scale rollout. This helps limit the impact of bugs or issues on the entire user base.

The first Dark Launching meetup was hosted in May, with a surprisingly large turnout of people interested in learning about dark launching itself. Dark launching is a best practice used by companies like Facebook to launch new features "dark" (off), then slowly light up (turn on) features for different users. The purpose of this approach is to mitigate the risk of bad releases and allow for easy turning off of features if issues are found. Despite thorough testing, real-world scenarios can still lead to unforeseen problems, making dark launching a valuable tool in ensuring user satisfaction. The meetup group plans to share more stories and lessons learned at future events.