Re: [Tracker] Tracker and Zeitgeist

[Jamie McCracken <jamie mccrack googlemail com>]

i dont see it that way

I want to create a tracker powered file manager and I need to have file
history dates so it has to be independent of zeitgeist (im happy to just
index the gtk recent files file for this info so its probably less than
a 100 lines of code to do and I can do it for you)

Im sure nokia will want to signal tracker when music files are played so
having an Events dbus api in tracker should be desirable. I also think
event notifications should be free desktop based so any app can fire off
an event using say org.freedesktop.Events on dbus

In general there is a lot of talk around zeitgeist so its not clear what
they will implement but there is plenty of time for that -6 months to
beta and 12 months to gnome 3 inclusion. 

One thing im sure about is any indexing being done in zeitgeost should
be replicated in tracker - I know zeitgeist is watching firefox sqlite
db for changes in bookmarks and web history so we should do the same.
After all if we are not indexing useful content then thats a bug surely?

jamie

On Sun, 2009-07-12 at 14:10 +0300, Ivan Frade wrote:
> Hi, 
>
> I think this topic deserves a new thread. It is not really related
> with 0.7 release plans.
>
> Zeitgeist is _not_ a wrapper around tracker. It is a different daemon,
> which will receive tons of events from different applications like:
>
>  open (app1, doc1,...)
>  open (app2, doc2, ...)
>  alt_tab (doc1, doc2, ...)
>  alt_tab (doc2, doc1, ...)
>
>  This information is not in tracker, but in the zeitgeist DB. Then the
> zeitgeist engine will read that, run some fancy algorithms and do the
> dark magic to generate something like:
>
>  doc1 zeitgeist:related doc2
>
>  This triplet goes into tracker (In this sense, zeitgeist is a
> "miner"). What we will have into tracker is the processed information
> and not a wild bunch of events with no meaning. Tracker is a metadata
> storage for applications, and it is not intended to be an efficient
> "data mining" backend! The "data mining" is done by external processes
> and only directly useful things are stored into tracker.
>
>  So the idea is:
>
> 1) Applications store events in zeitgeist daemon
> 2) Every now and then zeitgeist daemon push into tracker relations
> between docs, applications, tags, bookmarks, and so on.
> 3) Applications (including zeitgeist UI) use tracker to retrieve
> information
>
> On the tracker side, we just wait for zeitgeist developers to specify
> what properties do they need in the ontology.
>
> This is my understanding after some conversations with Seif and
> Federico about tracker and zeitgeist. They look happy with it, and  It
> sounds great to me. 
>
>  Comments, suggestions? anything obviously wrong with this approach?
>
> Ivan
>
> _______________________________________________
> tracker-list mailing list
> tracker-list gnome org
> http://mail.gnome.org/mailman/listinfo/tracker-list