There’s been a lot of posts written recently on the web about the state of tooling in front-end development and the opinion that many share that the environment has become overwhelming both for the beginner developer and the more experienced developer.
At Future of Web Apps 2015 I watched a really interesting talk from Peter-Paul Koch in which he argued that our tooling problem has become an epidemic; and that we should stop encouraging the creation of new tools and libraries. After my talk at FOWA, in which I demoed building applications using jspm, SystemJS and Babel, I had an attendee question if the addition of jspm was warranted. It’s a perfectly reasonable question and one that got me thinking.
I explained to the attendee that I use jspm because it solves a problem that I don’t want to have to deal with. In this case, jspm lets me install 3rd party modules from npm and GitHub without me having to deal with configuration or any form of additional build tool, and it also provides the bundling functionality when it comes to deploying my application. Yes, I pay the small overhead of adding another tool, but I can justify it.
The problem and confusion comes from using tools just because they exist or because they’re the “new shiny” toy that has come along. If you’re adding a tool to your workflow just because someone on the internet said you should, you’re going to hit difficulties. You need to strenuously vet tools before deciding that you’re happy to have them as part of your set up. The way you would do this is by using a tool on a project ultimately, but not without doing some research first. Most projects will provide examples of using them and you should try to marry those up with your project. If you find yourself struggling to explain to a colleague why you think this tool will be a good addition, it’s likely that it isn’t. Don’t force a tool upon an application that it’s not suited for. This isn’t just true for package managers like jspm, but frameworks, polyfills and any form of 3rd party item you might include.
Additional criteria you might check to decide if a tool is right for you and your project might be:
You should also forget about trying to “keep up” with the web as more and more tools come along. Tim Kadlec talks about this in his blog post “The Fallacy of Keeping Up”, which I recommend reading. It’s just not worth your time and effort to try every tool as they come out and chop and change. Recently someone asked me if they should leave Grunt to try something else like Gulp, and I asked them what problems they were having with Grunt. They weren’t having any, but this person had been told by many that they should consider swapping over. Yes, it’s true that if you’re relying on a tool that’s not been maintained for years you might consider moving over, but don’t let that decision be a knee jerk reaction. We have a skewed view in this industry; an “unmaintained” project on GitHub is one that’s not had a commit in three months. Think longer term, and don’t be afraid to stick to the tools you trust. You should end up with a tried and trusted toolset that you rely on time after time. For me and the projects I work on that’s jspm along with ESLint and a couple of others. For you it might be Grunt, JSHint and CoffeeScript. It doesn’t matter, as long as you can justify each of them to me and spend the time to build up your knowledge and understanding of them.
As we move more and more functionality to the client, part of the trade off is that we end up with a more complicated tooling set up. If you ever work on a server side language you’ll meet a bunch of tools that have grown out of the requirement to tackle the requirements of authoring and working with a large application. Traditionally in the past we’ve never had to deal with that, and hence there’s never been much of a need for a complicated tool chain. Now we’re building full, stateful applications in the client, it’s inevitable that we’ll need some extra tooling to help us, and we shouldn’t be afraid of that.
I don’t think that we’re creating too many tools - not only is it great to have a variety, it also encourages competition and improvement. ReactJS is a great example of this; since its introduction and adoption other frameworks like EmberJS have embraced React’s virtual DOM model, and others will follow suit too.
What’s more interesting to me is the type of tools we’re creating. I’m a big believer of the Unix philosophy: each tool should do one thing, and do it well. This is definitely something we can take into our approach with web development tools. Every individual tool that you add to your project should be able to have its functionality summed up succinctly. We shouldn’t look to use or to create tools that attempt to do everything. This is partly the reason ReactJS has been adopted so quickly - its surface area is very small, and the functionality it provides is well defined. This is partly the reason I’ve stopped using task runners like Gulp or Grunt, and instead prefer to install many smaller modules that each provide a small piece of the puzzle. Of course, I could use Gulp given that all its functionality comes from plugins, but I don’t need that additional complexity. You might, and if you do, that’s great. I’m not saying there’s anything wrong with Gulp, larger frameworks like Angular, or tools that provide a range of features. As we go forwards though we should make a concerted effort to build tools with smaller surface areas. The benefits of this are huge:
I hope that this article provides some food for thought, and I’d really love to hear what you think. Please feel free to drop me a tweet, I’d really like to discuss the issues of tooling and what we can do going forward to ensure we head in the right direction.
Thanks to Alex Young, Shane Hudson, Adam Onishi, Ruth John and Peter Müller for their review of this blog post.
If you enjoyed this article please feel free to share it on Twitter.