What a difference a year makes. Astute readers of this blog may remember my sitting on the Node.js fence in the fall of 2013. I eventually jumped off the fence thanks to the wave of presentations at the last year’s NodeSummit indicating that Node.js was ready. During the ensuing year I have become a vocal supporter of Node.js-based micro-services for fun and profit. It is therefore with great anticipation that I flew to San Francisco earlier this week to attend this year’s NodeSummit.
One of the immediate delights of my visit was the fact that unlike last year, IBM’s presence was huge (‘platinum sponsor’ big enough for you?). There have been projects in IBM dabbling in Node.js, and IBM PaaS Bluemix has a first class support for running Node apps. More recently though, Node.js has become a go-to technology for every new IBM project I have seen recently, particularly cloud-hosted (and most new projects fall into that category these days).
The theme of the last year was the claim that ‘Node.js is ready for the enterprise’ (once the Walmart memory leak has been fixed). The mood was one of hope and excitement, with a tinge of whistling in the dark. During the ensuing year, the pilot projects have wrapped up, risk-awerse enterprises have convinced themselves that Node.js is just fine for production, and there is no need to convince anybody any more. This year, we moved on to the details and the logistics of doing it at a massive scale.
Of course, the highlight of the conference was the announcement of the formation of the Node.js Foundation, taking over stewartship of Node from its sole corporate sponsor Joyent (Joyent continues to be the member of the Foundation). It is left to be seen if this move will heal the rift in the community caused by the fork of Node into io.js, but it does point at the maturation of the platform and addresses some of the key complaints regarding the control over its evolution.
Another long-awaited news was the shipping of Node.js 0.12. Exactly a year ago when I attended Node Day organized by PayPal, then and still current Node.js team lead TJ announced that the shipping of 0.12 was ‘immanent’. Well, this was the longest ‘immanent’ anything in the recent history, if that word would to retain any meaning. On the positive side, the shipping of 0.12 followed the passage of all the test case suites for all the supported platforms, turning this into the new quality benchmark to satisfy for all future releases (and the coveted 1.0). It also underlines the difficulty that Node.js platform now finds itself in, being built into the core of more and more production deployments, and under a lot of pressure to evolve the platform while maintaining quality.
One of my favourite talks was by Fred Schott from Box about the realities of running Node.js in a real-world situation. It was heart-warming to follow performance tweaks all the way from mediocre to fast. It is a cautionary tale that, while Node.js is a wonderful platform with a potential for great performance, careless out of the box apps may disappoint. Like any other platform, it requires tuning, profiling and (what a concept) a certain level of mastery. You cannot just copy a ‘Hello, World’ app from Stack Overflow and expect server-melting speeds.
Another personal highlight was that the old ‘server vs client’ rendering debate is alive and well. It was amusing for me to hear two young PayPal team members talk about ditching Node/Dust combo (a normal PayPal staple) for client side rendering using Angular. In my chat with them they revealed that they managed to wrestle decent performance from Angular on mobile only after a lot of work (see my comment about Box’s experience above). Funnily enough, in a later talk Peter Marton shared tricks on isomorphic Node.js apps, where the same template is reused on both sides of the divide (my preferred technique as well). It just shows that there is no right or wrong approach – whatever works in your particular situation.
If I had one complaint about the two days I spent at the conference, it was that it was somewhat light on hard-core technical content. The focus was on high-level panels, ‘this is how we migrated from the monolith to the Node micro-services’ talks and ‘look at all the boxes in our architecture’. Now that we have all congratulated ourselves for the foresight to support Node.js and turn it into such a world-class platform to write modern systems, we should probably take off our celebratory hats and dig into the details of hard core, production Node.
All in all, I enjoyed the summit but left with mixed emotions. Last year I was fully aware of doing familiar things in a new platform, and the most mundane tasks (‘look ma, I am setting a cookie using Node.js!’) felt new and exciting. I think we are leaving that ‘new car smell’ age and entering the period where Node.js will just dissolve into the background – become like air or water, and we will focus on ‘what’ we are doing, rather than the fact we are doing it in Node.js. It may become like (gasp) Java at some point in the future.
For some developers (see my post on another TJ) there is no fun in that (look, ma, I am setting a cookie using Go!), and restless Node.js committers eager to move fast and break things are already busy in the io.js repository. I on the other hand just want to build awesome apps. Node.js is already a great platform for me and as long as it remains stable, reasonably bug free and supported, I am happy. I am sure I share this with many enterprise Node.js enthusiasts.
© Dejan Glozic, 2015
Leave a Reply