Pierre Chapuis

We use nginx because it is a great frontal Web server by itself, but also because we rely on the Nchan module for server push.

We use AWS mostly because Amazon RDS for PostgreSQL is a very good and cheap solution to get a managed PostgreSQL.

We also use Amazon EC2 for the servers, with the Arch Linux images from Uplink Labs. We chose Arch Linux because of its up-to-date packages, especially for Python. We use Pacman to package and deploy our services.

We use TypeScript because it is the native language of Angular applications, but also because types helps prevent errors at compile time.

Unlike our frontend, we chose Flask, a microframework, for our backend. We use it with Python 3 and Gunicorn.

One of the reasons was that I have significant experience with this framework. However, it also was a rather straightforward choice given that our backend almost only serves REST APIs, and that most of the work is talking to the database with SQLAlchemy .

We could have gone with something like Hug but it is kind of early. We might revisit that decision for new services later on.

We chose Angular 2+ because our product has a rather complex back-office, most of which is visible only to our employees and contractors.

The "full featured" approach of Angular as well as the way it deals with the data model suited this business case well. We also use the Reactive Forms module a lot, as well as Clarity (an Open Source design system by VMWare) for the internal parts of the product. We also intend to use Angular's offline features.

For user-facing parts of the product we are now considering implementing views as Web Components with Stencil. We use Sass (SCSS syntax) for styling and TSLint to enforce a style guide close to Angular's default.