Re: [Tracker] Newbie question

[Sam Thursfield <ssssam gmail com>]

Hi Don

Welcome to the Tracker project !

On Tue, Jun 7, 2016 at 5:34 PM, Don Doerner <Don Doerner quantum com> wrote:
> Just joined the list, need to learn about tracker in a multiuser filesystem.
> Can anyone recommend (or better point me at) discussions of this topic in
> the archives?  I am specifically interesting in discussions regarding the
> partitioning of index data – I’d like to be able to specify a path (e.g., a
> home directory), index the files below that path and make that information
> available to a specific user (continuing the example, the owner of that home
> directory).

The list archives are searchable through https://mail.gnome.org/, if
you include the word "tracker" in your search terms you'll get mostly
results from tracker-list (there may be a better way to limit results,
I don't know).

The model of Tracker is to run per-user, and store its database inside
the user's home directory. So Tracker doesn't have much support for
partitioning data within a single database. On a multi-user system,
each user has their own tracker-store and tracker-miner-fs processes
running inside their desktop session.

There are downsides to this model, but it has a big advantage of being
simple and secure (secure in that it takes advantage of the per-user
security model the OS provides, rather than trying to reimplement the
same thing itself).

That said, when you add a statement (an RDF triple) you can specify
which "graph" it gets stored in. So there is a way to separate data
within one database. This feature is referred to as "named graphs"
usually. It's not something that can be relied upon for *hiding* data,
because all the data still ends up in the same SQLite database which
can be read with the 'sqlite3' program, but it can be used for
partitioning data.

Sam