A peculiar issue seems to be happening since 4.9: some of my files have ridiculously long lists of timestamp values in their 'modified' 'date' and 'annotation' metadata fields. Some investigation seems to indicate that they are associated with KTorrent. I initially figured that as the file was being gradually downloaded nepomuk was repeatedly reindexing the updated file and annotating the process, so I used the KTorrent feature that lets me download to a temporary folder, and just checked off that folder from strigi's index list. That didn't help, and further experimenting seems to reveal that *seeding* was causing the problem: freshly downloaded files were clean, while files that have uploaded some data to the torrent network were plagued with the huge lists.

I found a reviewboard patch (https://git.reviewboard.kde.org/r/103794/#review10102) regarding strigi reindexing files that are seeding with KTorrent and some other stuff. As I understood from the comments there, strigi does need to check the file after it is closed, but won't reindex it if nothing has changed. This sounds reasonable, but 4.9 onwards strigi seems to be annotating the file with a timestamp after every check whether or not the file has really changed.

It can get really annoying: the worst affected of my files have such monstrously huge lists that they lag dolphin for 10+ seconds loading them in the information panel, and when it finally finishes loading I can't even scroll the information panel without lagging it more. Can anyone confirm this happening? It should probably be reported as a bug.

I am not sure, but this might be fixed in 4.9.1.
Please see this blog post, especially "Reindexing events" section.
http://vhanda.in/blog/2012/09/nepomuk-and-kde-4.9.1/

That looks very promising I'll wait for 4.9.1 and hope it's done then. If I still see the issue then I'll ping again. In the meantime I'll just write a script to clean up my files manually and temporarily avoid seeding in ktorrent.

Try not to index directories with temporal or downloading files to not stress so much indexing process. Temporal files only generate noise and indexing partially downloaded files could cause indexing malfunctions.

Ignacio Serantes wrote:Try not to index directories with temporal or downloading files to not stress so much indexing process. Temporal files only generate noise and indexing partially downloaded files could cause indexing malfunctions.

Yeap I know. However this problem is with fully downloaded files that are uploading to the torrent network. I do downloads in a temporary folder that isn't indexed.