We actually initially wrote a lot of networking code in Kotlin but the complexities involved prompted us to try and compile NodeJS for Android and port over all the networking logic to Node and communicate with node over the Java Native Interface.

This turned out to be a great decision considering our battery usage fell by 40% and rate of development increased by a factor of 2.

As a small team, we wanted to pick the framework which allowed us to move quickly. There's no option better than Rails. Not having to solve the fundamentals means we can more quickly build our feature set. No other framework can beat ActiveRecord in terms of integration & ease-of use. To top it all of, there's a lot of attention paid to security in the framework, making almost everything safe-by-default.

Hey guys,

My backend set up is Prisma / GraphQL-Yoga at the moment, and I love it. It's so intuitive to learn and is really neat on the frontend too, however, there were a few gotchas when I was learning! Especially around understanding how it all pieces together (the stack). There isn't a great deal of information out there on exactly how to put into production my set up, which is a backend set up on a Digital Ocean droplet with Prisma/GraphQL Yoga in a Docker Container using Next & Apollo Client on the frontend somewhere else. It's such a niche subject, so I bet only a few hundred people have got a website with this stack in production. Anyway, I wrote a blog post to help those who might need help understanding it. Here it is, hope it helps!

This basically came down to two things: performance on compute-heavy tasks and a need for good tooling. We used to have a Meteor based Node.js application which worked great for RAD and getting a working prototype in a short time, but we felt pains trying to scale it, especially when doing anything involving crunching data, which Node sucks at. We also had bad experience with tooling support for doing large scale refactorings in Javascript compared to the best-in-class tools available for Java (IntelliJ). Given the heavy domain and very involved logic we wanted good tooling support to be able to do great refactorings that are just not possible in Javascript. Java is an old warhorse, but it performs fantastically and we have not regretted going down this route, avoiding "enterprise" smells and going as lightweight as we can, using Jdbi instead of Persistence API, a homegrown Actor Model library for massive concurrency, etc ...

I was researching multiple high performance, concurent//parallel languages for the needs of authentication and authorization server, to be built on microservice architecture and Linux OS. Node.js with its asynchronous behavior and event loop suits the case best. Python Django & Flash turns to be slower and .NET Core & Framework wasn't the best choice for the Linux environment at the time (summer 2018).

I also tested Go lang and Rust, although they didn't meet the quick prototyping criteria as both languages are young and lacking libraries or battle-tested ORM.