What technology stack would I choose, if I had to develop a web application completely from scratch? That’s the question this article will cover.
First of all: by saying web application I’m referring to something between a plain static HTML page and an entire Facebook. Basically, an application that fulfills a certain domain of tasks for the user and that requires the usual features like user management, a backend database, multiple UI views and controls, etc. The size of application I’m thinking of could be a browser-based chat app, password-manager or something similar. Neither too simple, nor too complex.
Of course, the technology choice depends on the concrete project requirements to a certain extent, but not completely. Consequently, a new project is always a chance to try something new. ThoughtWorks just published their new technology radar for 2016, where they separate into categories adopt, trial, assess and hold. Of course, hold-techs are not an option for new projects and I actually pretty much agree with their views on what has to be in the hold category. Adopt basically are things that are modern, but also well-established enough to avoid too much risk. Trial-techs are more experimental and assess are the latest fancy s***, so to say. Since I’m extremely eager to try out new things, my stack would probably mostly consist of technologies from the last category. So what would my stack now look like? Actually, I couldn’t decide on one stack, but set up two: the fancy one and the super fancy one. Additionally, I define their intersection as the base stack, which consists of fundamental tools etc. that both have in common.
First of all, I’d use Git for version control, Visual Studio Code as code editor and GitLab for repository hosting and as build server. If I didn’t had to implement user management myself, I’d pick Auth0 for that. For deployment, I’d use containers with Docker on DigitalOcean machines and if I needed multiple instances, Rancher would help me to manage them. As reverse proxy in front of the backend I’d choose nginx since it’s extremely efficient, performant and has HTTP/2.0 support. For bundling, Webpack would be my choice and task automation would be done using plain npm scripts. For styling the UI, I’d simply use Bootstrap 4 and SCSS.
This stack differs from the super fancy stack in a few points. A key point is that it would not use GraphQL, but a good old REST API. This API would be written in Go, since I like the language - especially its efficiency and its good suitability for web development. More precisely, I’d use the Iris framework. I’ve read the documentation and it looked incredibly powerful to me (in terms of both functionality and perfomance). (EDIT: It’s not what it seems! Please see my comment below!). For the frontend I’m balancing between Angular 2 and Aurelia. Angular 2 is guaranteed to work for any potential case, is extremely powerful and has great community- and library support. However, Aurelia look promising, too, and probably is even more clear and less boilerplate code. Consequently, I’d give it a try. But if having to go mobile, I’d still favor Angular 2, since it perfectly aligns with Ionic 2.
Two other options, which look really interesting to me are Meteor and HorizonJS. However, I’m not sure, if it’s a good idea to commit to only one comprehensive framework through the full stack.
So these are my two alternative ways through the webdev jungle - btw, this good article describes another one, especially for newcomer web developers. Sorry, that I haven’t justified all choices. Actually, as you probably know, if you’re a developer, subjective views like these often aren’t even based on pure rational considerations, but are rather emotional and spontaneous.
Please feel free to give me feedback on my tech stack of choice!
EDIT: Another framework I’d really like to try out is InfernoJS, because it claims to be extremely lightweight and performant. However, before using Inferno, one should probably be familiar with React, since it uses very similar concepts and syntaxes.
EDIT 2: After having read this article and having done some further research on the Iris framework I really have to retract my above statement that I’d use it as a web backend. While it looks nice on paper, after diving a little deeper I really have to admit that it’d be morally tenable to support the authors of that project. So please forget about Iris and take a look at Beego instead.