This was my second year attending Velocity in Amsterdam, and as with 2015 I shared my time between the Akamai sponsor booth and the talks themselves. Here are a few of my hot takes from this year's conference.
HTTP/2 Data in the Wild
Understandably, there was a lot of buzz around HTTP/2 last year, and this year we're starting to see some decent data on the effect of enabling the protocol. Unfortunately, it's not all good news and as with much in web performance, there's nuance in the detail.
Fortunately, there's a number of case studies and tools to help understand how best to implement H/2. Colin Bendell announced his Can I Push? browser H/2 feature support checker and Should I Push? H/2 push opportunity tool during his talk (slides: The Promise of Push). As somewhat of a counterpoint, Yoav Weiss discussed the use of preconnect and preloading to address the same issue that H/2 push is targeting - making the use of spare bandwidth during page loads (see: On Preloads and Preloaders) .
Michael Gooding and Javier Garza presented some highly useful real-world data that ran often ran counter to some of the prescribed H/2 best practises. For example, unbundling your assets into separate files isn't necessarily a good idea, so it's worth considering resource bundling and only separating assets that change frequently. I'd highly recommend checking out the slides here: Real world experiences with HTTP/2.
Images - and what to do with them
Much of the chatter at the Akamai booth was around images and their significant implication on web performance. Naturally a lot of this was inspired by our Image Manager demos and our giveaway of the High Performance Images book. However, it was a topic that touched a nerve and I had a lot of interesting conversations around how images are being managed (or not) at the moment, one that stuck out covered an application that loaded over 7mb of images per page.
Although everyone I spoke to seemed to understand the impact of images on web performance, not many had a handle on how to minimise that impact - and do it at scale. For more information on how Akamai's Image Manager could help, see: Image Manager | Automate Image Optimisation and Delivery.
Some of my favourite talks reach outside the subjects I'd normally consider in my day-to-day work. Steven Shorrock discussed the gap between Work as imagined and work as done - and the dangers of implementing controls around work based on an (often incorrect) view of how work is actually done - which can lead to excessive monitoring, over-proceduralisation and other such "wallpaper" solutions. There were interesting parallels between this and the follow-on keynote by James Duncan on the widening gap between the bleeding edge of technology, and the day-to-day work of technology practitioners. Beyond those gaps, the fear of failure in large organisations often leads to the rejection of ideas - leaving small organisations with a culture that's more accepting of risk to incubate those small ideas, and turn them into great ones. No matter the size of your team, department or company - it's important build a culture that understands how work is done, supports those practises, and enables ideas to fail or flourish.
Sarah Wells offered a fascinating insight into how the tenets of Nudge Theory were used at the Financial Times to design services and systems, and guide teams to adopt them. The easy adoption of cloud services can result in an unsupportable mix of different services, and Sarah walked us through the steps to make sure that the services were designed in a way to make them fit for purpose, attractive to use and easy to deploy.
I've found the psychology of web performance fascinating and I was interested to see Denys Mishunov present some of the subjects he'd previously covered in his 3-part Smashing Magazine piece. Naturally the three-part article offered a lot more detail than the session, but nevertheless it was interesting to here Denys cover the psychology that determines how performance is actually perceived by the brain - and adds an extra factor to consider when designing for performance. Although the slides for the session aren't online, it's well worth reading the three-parter (especially if you've ever wondered why there are always mirrors in elevators).
The topics at Velocity have become increasingly DevOps and systems focussed, and that's why it wasn't a great surprise to here that the web-performance track would be moved to the Fluent conference next year, leaving Velocity with a true systems engineering message. Hopefully the change manages to maintain a diverse set of topics while continuing to push web performance forward.