GNOME Bugzilla – Bug 680808
Disable crawler by default
Last modified: 2013-07-25 12:35:31 UTC
The crawler should be disabled by default (as requested in bug 613256) and be driven "on-demand" by either applications (bug 613252) or by a session-wide utility (which could take into account the state of the machine such as power status, usage, logged in users) to reduce interference. A session-wide daemon (whether session manager, or a long-running daemon) would know when the best time is for scheduling indexing, and could make indexing stop when particular events occur (manually requested backups, user activity, etc.).
(In reply to comment #0) > The crawler should be disabled by default (as requested in bug 613256) and be > driven "on-demand" by either applications (bug 613252) There are generally 2 models people want. 1. App driven 2. Event driven We support a combination of the two. The crawler can be disabled by default if you patch the tracker-miner-fs.desktop file to have X-GNOME-Autostart-enabled=false. I don't think it makes sense to switch to either one or the other model exclusively. Model #1 (i.e. app driven) makes sense ONLY when you have a controlled environment. GNOME is not a controlled environment. The N9 phone Nokia produced *is*. So I have these *image* applications on my machine: - Inkscape - GIMP - Shotwell - F-Spot - Dia - EOG - SimpleScan - Evince and some others which I am playing with right now. Are you telling me we should patch ALL those applications *JUST* for image support in Tracker? Plus, we've not even got on to the myriad of music and other document type applications out there yet. It's just not a manageable solution. > or by a session-wide > utility (which could take into account the state of the machine such as power > status, usage, logged in users) to reduce interference. On a session-wide utility: I think this is a good idea sure. Tracker does need some work to take advantage of idle times only (i.e. at lunch time when a user steps away). Right now SCHED_IDLE is the only fallback we have close to that right now. On power status: This is configurable by the user (i.e. don't index on battery). However, this makes tracker quite useless. Our credibility is our data being up to date. Would you use Google if it found out of date results? On usage, logged in users: I think this is a good point. I could argue that since Tracker is not started until a user logs in, it's moot. Also, if another user is doing something, the other Tracker processes won't be doing anything but idling will they. Do you have evidence that Tracker is an issue for a multi-user system? On interference: I have to say, I also don't see any interference from Tracker on my machines. Do you have evidence to suggest otherwise? > A session-wide daemon (whether session manager, or a long-running daemon) would > know when the best time is for scheduling indexing, and could make indexing > stop when particular events occur (manually requested backups, user activity, > etc.). There are many ways to do this. Nokia did it with cgroups. We do various checks around disk use, battery availability and of course we try to reduce CPU/disk hogging. A daemon might be useful here to avoid us doing what some other GNOME apps may also be doing, but it's not within the scope of the Tracker project to produce that daemon is it? or is it? -- This bug is really about more than disabling the crawler by default. I would prefer it if we kept bugs as simple as possible instead of discussing many issues in one report. For now, I will allow some discussion, but based on the bug title, I am most likely to close this as WONTFIX. It's possible to do what the bug report suggests for each distro that cares to apply a 1 line patch to their packaging, but I don't believe Tracker is useful out of the box this way at all.