Also regards to The Register's 'trendy' accusation: GDS chooses what works over what's trendy for everything critical: their apps work without AJAX and CSS too.
Also regards to The Register's 'trendy' accusation: GDS chooses what works over what's trendy for everything critical: their apps work without AJAX and CSS too.
Absolutely. The vast majority of the site is a Ruby on Rails content management system, as the reader will find out if they visit the github. There's a few shiny bits that are Go and MongoDB and whatever else (primarily things like load balancers and API endpoints and dynamic metadata services), but most of it is a stack which is, frankly, mostly stable and "enterprise" by now.
Their primary alternative to the system they've built would've been a huge, proprietary content management system - Microsoft SharePoint with a lot of customisation comes to mind, and likely would've cost them more, as well as reducing the pool of developers with experience.
Erm no. One GDS project requires end users to connect to a number of devices over USB from their Windows PCs. The logical choice is to write a .NET app that does it.
GDS chose to write the USB interop code in Java, wrap that up in a Windows Service, install a web server on the PC, write an HTML/CSS/JavaScript app that runs in the web server, which calls the Windows Service and talks to the USB device.
That sort of thing (the above is not an isolated example) happens at GDS for a number of reasons - one is what the Register alluded to: they're good at webs sites, so they try to shoe-horn everything into the browser. Another is that they're despotically obsessive about "user need", so don't understand that user needs need to be balanced with business needs (reduce call centre costs), system needs (we've invested x in Windows, for better or worse), and operational needs (supporting a single, self-configuring .NET .exe on 2,000 desktops through WSUS is a lot cheaper and easier than installing, configuring and maintaining a Windows service, a web server and an HTML application on 2,000 desktops).
This in the context of developers finish the system, leave, and ops don't have the skills to fix a Ruby/node/whatever tech issue.
While none of the options that work are technically wrong, it might be worth considering the differences between the best solution design, and a successful one.