[gumby]

I don't understand the implementation. They already have a background process running (on my Mac it shows an icon in the menu bar) and that application talks to the network. Why use a second process? I can understand spawning a process to actually do an update once one has been downloaded. So what am I missing?

(and since this is HN: as for the updating, fine. It's no different from visiting a web site and having the pages served by different revisions of the back end, and it's not like it has a complex UI that could cause user confusion. I assume DB doesn't update all that often anyway).