Today I had a closer look at localisation, and how to automate it based on their frequency of use per workstation.
Most of my notes have gone into their corresponding issue on GitHub, so I'll link to those here.
The first step of the journey was figuring out whether we could leverage the already existing memcached setup to spy on commonly used packages per machine. We could then use this information to automatically localise packages where necessary, being as efficient as possible with disk space.
However that route didn't pan out.. There was a feature that would have gotten us there, but the developer kindly informed me that it has gone bust. :(
In awaiting his reply, I took to designing the mechanism itself and how it would work once we have the information, which you can find in the Design Document linked above.
I'll implement the MVP of the localiastion feature, and take it for a spin.