Re: Proposed module: tracker

[Jamie McCracken <jamiemcc blueyonder co uk>]

Iain * wrote:
> On 1/8/07, Vincent Untz <vuntz gnome org> wrote:
> > Information about tracker:
> > http://www.gnome.org/projects/tracker/
>
> I object to tracker not on any technical merits, but as a project I
> don't think we have a concrete idea of what we want to use it for, I
> feel that its getting proposed because its a piece of cool technology
> and we don't want to miss out on not having it to compete with other
> desktops.

T-S-T siginificantly imporves G-S-T and thats why its proposed. No one 
has said why thats a bad thing (although ebassi asked some questions why 
its better which I have already answered and will not repeat here)

I have made it clear that it was T-S-T and "tracker the indexer" that I 
was proposing. Objecting to "tracker the database" which some are doing 
is a bit out of scope.

The database side of things would only come into play if maintainers 
wanted to use it and only if they were happy with it.

> I would be scared that it would be added to GNOME before we work out
> our actual requirements for it, and without the requirements how can
> we know that tracker solves them?
>
> Unfortunatly, the only way to know this is for applications to try to
> use it and find out what is missing, what it does well and what it
> doesn't do well. This is, to me, the solution for the chicken/egg
> race. The applications use it to find out if it does what is needed
> and then if it does, we add it to GNOME.

Minor integrations like into Gedit could be done conditionally but major 
stuff like the epiphany bookmarks would not be practical without it 
being at least an approved dependency as it would involve big changes to 
the Epiphany UI as well as the bookmarks storage.

The ephiphany integration would be great cause it would really show off 
all of tracker's capabilities (it would expose all the cool stuff in 
tracker like effortless search, stats for a timeline widget, extensible 
metadata, first class object storage etc)

It would be nice, if tracker fails to get in the desktop, to allow 
tracker to be an approved dependency to make the above possible and also 
prove its worth. A branch in Epihpany would be a poor second considering 
the lack of guarantees, amount of work, maintenance of branch and the 
possibility of it ending as a fork of Epiphany if it does not get merged.

I understand why gnome needs approved dependencies but they need to make 
sure it does not become a barrier to adoption of new technologies too.

> Another worry would be that really an indexer would ideally be desktop
> agnostic and I'd like to see KDE and GNOME share an indexer/metadata
> store/whatever tracker is, because I don't want two indexers indexing
> my data for each desktop. Is there any moves to getting tracker used
> in KDE as well?

yes we have a QT gui written. The tracker daemon is desktop agnostic and 
XFCE people are highly interested in it too. KDE also has strigii though 
but work is on going in freedesktop land to create a unified dbus 
interface for indexers like tracker, beagle and strigii.

> (Plus I note that jeff/ebassi's concerns about what "tracker" is
> haven't really been answered)

If you know what Beagle is and you know what Evolution Data Server is 
then tracker is the sum of the above two (and really its not more 
complicated than that). The EDS like functionality is more generic 
(RDF/semantic web based) in tracker's database so can handle more than 
contacts, appointments etc.

the advantage of combining the two is greater efficiency, avoids 
duplication of data and allows it to be more extensible and customisable 
to users/apps needs.


--
Mr Jamie McCracken
http://jamiemcc.livejournal.com/