Comment 30 for bug 646724

Revision history for this message
Mikkel Kamstrup Erlandsen (kamstrup) wrote :

The reason that we are not using tools like Tracker and Beagle is that we didn't find them "reasonable performance-wise", nor providing the quality of results, or fine control of queries we wanted. And working "reasonably well" is just not good enough in my book.

That said, there's nothing saying we must stick with ZG, it's just that it is an easy practical solution.

My personal check list for an indexing tool is something like:

 a) Doesn't crawl you home dir on login (to install inotify watches or check for new stuff)
 b) High quality query relevancy ranking
 c) Able to sort by usage frequency or recency
 d) (wishlist: direct access to index api to inspect term stats etc without the need for DBus roundtrips)

Solving a) requires something like btrfs or fsnotify. To my knowledge both Tracker and Beagle rely on inotify (possibly indirectly via GIO). And Tracker fell short on b) + c) + d) last I checked. Solving b), c) , and d) seems to inevitably require something like Zeitgeist combined with Xapian or Lucene (much like we do now :-)).

My currently preferred solution would be to:

 i) Extend ZG with a new range of data providers (music, u1, telepathy, contacts, etc)
 ii) Write a Nautilus extension that makes ZG index all files and folders you ever see in the file browser (this may require some filering logic in order not to spam your log too badly)