Re: [Tracker] indexing ~/. dirs?

[mike <mhardy_mail yahoo com>]

--- Jamie McCracken <jamiemcc blueyonder co uk> wrote:

> On Tue, 2008-04-08 at 10:11 -0700, mike wrote:
> > mike wrote:
> > > > I haven't been able to figure out how to get
> > > > directories such as ~/.dir_name into tracker
> to
> > have
> > > > tracker index what's there (I've added the
> paths
> > in
> > > > cfg file (separated/terminated correctly),
> didn't
> > do
> > > > anything; re-indexed, still didn't do
> anything) -
> > is
> > > > this possible?  I ran -v 2 and none of those
> dirs
> > show
> > > > up in the log.
> > >
> > > Running it with -v 3 might tell you more.
> > > I have a feeling we don't search .* directories
> on
> > purpose.
> > > src/trackerd/tracker-process-files.c should tell
> you
> > more.
> > >
> >
> > Yes, I can see that by default it does not check
> any
> > ~/. dirs (other than ~/.evo* and ~/.gaim) but I
> have
> > explicitly added other ~/. to tracker cfg
> > ("WatchDirectoryRoots") and it seems to be
> ignoring
> > the additions - is that intentional too?  I
> realize I
> > can probably just make links to them in ~/. but I
> > don't want a mess of ~/ links.  
> >
> > Given that many ~/. may be cache dirs or others
> not
> > suitable for searching, I can understand why the
> > default is to ignore them but even when explicitly
> > added to the cfg?
> >
> > In r0.6.6 I find no
> > src/trackerd/tracker-process-files.c nor any
> *process*
> > - is this something unreleased or just the wrong
> name?
> >
>
> it intentionally ignores hidden files and
> directories as we assume they
> are either worthless data or belong to some other
> non-file service like
> Conversations, emails etc
>
> We would require some justification for allowing
> explicit .dir indexing
>
> jamie

Fair enough but just to be sure, we're talking about
two different things, correct? 1) by default tracker
ignores ~/.* and 2) tracker ignores ~/.* even if added
to cfg file?

#1 seems not unreasonable to me but #2 does (in that I
explicitly add it to cfg but tracker ignores cfg
anyway).

What I use tracker for is searching for information
but that information isn't confined to emails or gaim
logs or unhidden files/dirs - could be browser data,
CD data, gconf/gnome settings, Palm PDA data, licq
logs, mime settings in ~/.mailcap, mplayer cfg, pan
newsreader cfg/data, xchat logs, pine mboxs, etc.

These (and other) dirs/files in no order are: ~/.X*,
~/.azureus/, ~/.bash*, ~/.csh*, ~/.bmp/, ~/.cddb/,
~/.config/, ~/.e/ (enlightenment), ~/.cups/,
~/.devilspie/, ~/.emacs*, ~/.galeon/ (browser),
~/.gconf/, ~/.gnome*, ~/.gnus*, ~/.gtk*, ~/.gxine/,
~/.jpilot/, ~/.licq/, ~/.local/, ~/.log*, ~/.mail*,
~/.mplayer/, ~/.nvidia-settings, ~/.pan/, ~/.pine*,
~/.procmail*, ~/.vmware/, ~/.xchat2, ~/.x*,
~/.xopenoffice (or ~/.ooo/), ~/.wine, etc.

Generally these are all entirely text-based; and I've
left out many dot files and other optional ~/. dirs
(such as ~/.elisp where 30M of my gnus usenet
newsreader data is saved so as of now, unsearchable).

But my point is that when I want to search for data,
settings, clues to computer problems, etc., because of
the convention of linux using so many ~/.*, that
relevant info isn't always going to be unhidden
files/dirs.  And in fact I'd argue there is a fair
amount of desirable info in these hidden files/dirs
that tracker is currently ignoring.  To be honest (and
I do like tracker), tracker loses much of its utility
for me when excluding entire sections of my computer's
data.

So I don't necessarily mind that tracker's default is
to ignore these, just that once they're added to cfg
that tracker honor its own settings.



      ____________________________________________________________________________________
You rock. That's why Blockbuster's offering you one month of Blockbuster Total Access, No Cost.  
http://tc.deals.yahoo.com/tc/blockbuster/text5.com