In 2013 the browser became the application. AngularJS 1.x made it practical to build rich, single-page interfaces that talked to a JSON API, and the pattern suited the dashboards, order screens and reporting tools Eziline was building. The year was spent splitting back ends from front ends and learning what that cost.
APIs first
Once the front end is a JavaScript app, the back end becomes an API. We restructured our Laravel work around REST endpoints returning JSON, with authentication tokens instead of sessions. It was more work up front, but the same API could then serve a web dashboard, a mobile app and a partner integration, which is exactly what clients started asking for a year or two later.
Angular in the dashboard
Two-way data binding, directives and dependency injection took some learning, and the early versions had performance cliffs with large tables. But for the screens that mattered, live inventory, point-of-sale terminals, patient lists, the responsiveness was a step change over page reloads. By the end of the year most new admin interfaces were Angular apps on a Bootstrap skin.
Mobile and design
- Native Android work continued alongside the web, and the first iOS projects were delivered for clients who needed both stores.
- Flat design replaced skeuomorphism in the design studio; iOS 7 made the change official in September.
- The multimedia team moved fully to HTML5 and video, with 2D animation work that still appears in our services today.
2013 also brought the first long-term maintenance contracts: clients whose systems were now central to their business and who wanted Eziline on call. That recurring relationship, rather than one-off builds, became the backbone of the company.
Looking ahead
The lesson of 2013 was that a modern application has at least three front ends, web, Android and iOS, and one API behind them, and that the company had to be organised accordingly. The plan for 2014 was to form a dedicated mobile team instead of borrowing web developers for app work, to adopt a consistent design system across the three front ends, and to fix the deployment problem that was now the main source of late nights: environments that behaved differently on each developer's machine and on each server. The Docker project that appeared in 2013 looked like the answer and we intended to try it early. We also wanted to turn the growing number of maintenance contracts into a proper support operation with a ticketing system, response targets and a rota, rather than whoever was free answering the phone.
