Bring back full-content indexing
Closed, WontfixPublic

Description

https://launchpad.net/ubuntu/+source/kubuntu-settings/1:18.04ubuntu6 added a Kubuntu Look & Feel package that implements a bunch of usability improvements that we agreed on, which is awesome!

However, one change snuck in that was not discussed and which I feel does not represent an improvement: disabling Baloo's full-content indexing by default. This change reduces the usefulness of KRunner and the Kicker Search. Full-content search is very useful for regular users, who can use it to find files if they remember something about the content even when they don't remember the file name. My whole family and extended circle of friends are non-engineer types who I would consider average users, and all of them make use of full-content indexing on their platform of choice to find files whose names they don't remember.

We should consider reverting this change and bringing back full-content indexing.

ngraham created this task.Mar 21 2018, 5:08 AM

-1

Most people I know, techie or not at all, disable such indexing where it exists and they are aware of it, so I dispute that conclusion for any claim of popularity.

Long standing CPU and resource hog. (many complaints over the years, including me)

Barely maintained or unmaintained upstream. Recent discussions about actually having a maintainer for it! Maybe if Michael Heidelbach does some more good work over the next few months as a new maintainer then it could be supportable again with full features, but I disagree that it is so for an LTS release (I'm not prepared to do so).

To be blunt, my actual feeling would be to get rid of it altogether. Ubuntu decided not to install or activate tracker - does a similar job and has similar issues - and support this for an LTS release. I don't think its really supportable for us either, but just disabling the full content indexing is a measure that leaves it there for people who want to 'opt-in', which is really what default should be for such thing.

Opensuse and upcoming Solus and some others set the same option, so being set to on is not ubiquitous or an obvious expectation.

Okay, so our substantive objections are that it's a resource hog. Do we have metrics or bug reports demonstrating this, and can we confirm that the problem disappears when Baloo is still on but only doing filename indexing? I personally have never experienced high CPU or memory usage caused by Baloo. My primary machine is a laptop and I get about 8 hours of battery life with it when doing real work, and I do measure the energy impact of various components.

I'd like to invite @michaelh to get his opinions on the matter, as he is effectively maintaining it along with @smithjd. I'd hardly call it unmaintained or barely maintained.

I'm not happy to say this: Switching off full-text indexing by default is a reasonable thing to do.
There are many bugs reporting baloo_file going berserk, specially when large text files are involved. So I think it's better to get full-text indexing out of the way for some time. Testing this is difficult. To mimick real-life one needs a lot of files and some really large ones. Also with only James and me contributing to baloo there is only little man power. And the list of crash reports is long ...

In short: Useful as it is, currently it is not feasable to provide this feature in a stable fashion.

ngraham closed this task as Wontfix.Mar 21 2018, 2:50 PM

Darn, straight from the horse's mouth.

Okay, let's leave it off for the LTS and consider bringing it back for non-LTS releases.

Darn, straight from the horse's mouth.

Okay, let's leave it off for the LTS and consider bringing it back for non-LTS releases.

Definitely. :)

ngraham added a comment.EditedMar 21 2018, 3:05 PM

Cool, sounds like a plan! Hopefully Baloo is in better shape by then anyway. Michael has been doing an amazing job of rehabilitating it so far.