System search design

Yesterday we had a call to discuss the future of shell search. We discussed various changes there, some of which can be implemented independently. However, there was also a proposal for the shell search architecture, which involves some interconnected changes. It went roughly like:

• Use a daemon to provide the shell with tracker search results
• Instead of using search providers for the core GNOME apps, have those apps put their data in the tracker database
• Give some of the tracker search results custom UI layouts in the shell (so we can give image results bigger thumbnails, and so on)
• When performing searches, delay querying the search providers, to give time for other search results to be provided first
• Once we have the search daemon, make it the basis of a more sophisticated search implementation:
  • Support text-based filters like type:pdf, type:doc, after:yesterday (possibly through a shared library)
  • Support unit conversions, so a search for 1km or 1 km returns 0.621371 miles

Let's use this ticket to develop this plan. If someone wanted to do a more detailed technical proposal, that would be cool.

@sthursfield @carlosg @chergert @fmuellner @bertob @jimmac @snwh @3v1n0 @dylanmccall @hadess