The State of Desktop Search on Linux


Desktop search is one of those techs that keeps coming back and never really sticking.  At least that’s how it seems to me.  Look at how giddy I was about Beagle back in 2006.  And I tried it and it was, generally, pretty awesome.  It really worked well.  It was like the speed of locate without having to wait until the database was updated at midnight.  And it could see into IMs, MP3 metadata, emails and office documents.  Now?  According the the official website it’s no longer under active development.  Perhaps that’s because they met all their goals.  And that’s fine, but pretty much everyone switched over to tracker around 2008.  I’m not quite sure why – perhaps all the anti-Mono hatred that went on.  Fedora doesn’t even ship with Tomboy or F-Spot anymore.  They’re there in the repos, but they aren’t the defaults.

But even tracker doesn’t appear to have quite the lustre it did two years ago.  I’m pretty sure when I re-installed Fedora when I went 64-bit that it didn’t have tracker installed.  I’m pretty sure when I checked yesterday it wasn’t installed.

I’ve been using KDE all week and I saw that NEPOMUK kept indexing my new music files when I bought music.  It was using strigi-libs, but strigi itself wasn’t installed.  It was also pretty much only scanning my music directory by default.  (there were some others, but I didn’t pay attention)  I told it scan my entire home directory.  Then I installed strigi.  The gui is pretty ugly compared to anything I’ve ever used for tracker.  And it didn’t seem to know anything about any of my files.  Dolphin used NEPOMUK to quickly find some audio-files but I couldn’t think of a test that would check that it was getting metadata rather than filenames since all the metadata is in my filenames.  Well, later on I was able to confirm that it is searching inside of files.  But it does’t seem to see everything just yet.

So why isn’t search going anywhere?  Was desktop search over-hyped? I remember being told that within a few years no one would care about folders because you’d be able to instantaneously find it.  And it wouldn’t matter if it was called 98q2iohnasioi8h.jpg because the search tool would read into the metadata and find that it was your cousin Earl.  But it doesn’t really seem to be happening.  And that really sucks because I know when I was using beagle I found it to be incredibly useful.

,

7 responses to “The State of Desktop Search on Linux”

  1. I think what happened is that people realized it’s really hard. 🙂

    NEPOMUK is still very active, AFAIK, though their approach is a bit more complex than just having a search tool; a lot of it is about contextual use of metadata and stuff.

    I don’t know what’s happening with Tracker, or what the integration plans for GNOME 3 are. Last I tried it could still quite easily get stuck on a bit of metadata it couldn’t chew, and stop working.

    • Yeah, the semantic stuff sounds awesome. It’s what I was most excited about when I read about KDE 4.0. I haven’t started working on it yet since i”m just getting back into KDE, but it could be pretty awesome. Looking at the NEPOMUK site, I see it’s a huge international thing and that Gnome is also involved – I think via Freedesktop.org. That’s be great instead of having two different indexes of my computer.

  2. Have tried a few of these tools over the years. Strigi, Tracker, Beagle, Windows Search. Probably some others. Without exception they have slowed my PC significantly, either all of the time, or on some regular and extremely inconvenient schedule, or at seemingly random intervals.

    I am not impressed by a tool that slows my PC all, or much, or the time just so that on the odd occasion I wish to search for something it might be found quickly.

    Better to use find, grep, exiv2 (to extract image metadata) and possibly locate.

    • Agreed that it shouldn’t cause much slowdown. I think they need to essentially do what the latest Norton Antivirus does – schedule that stuff for when the computer’s sitting idle