Re: User's preferred search tool

["Joe Shaw" <joe joeshaw org>]

Hi,

On 11/2/07, Denis Washington <dwashington gmx net> wrote:
> Why not use the xesam spec? We could just have one generic GNOME search
> tool that sends xesam queries and presents the search results,
> regardless of which search engine gave them us. This would require a
> xesam interface for basic file search, but that shouldn't be much of a
> problem.

I don't think anyone is in disagreement, but this is easier said than
done.  Beagle, for instance, doesn't use D-Bus for its standard IPC,
so it's not just a matter of implementing an additional D-Bus
interface.  There are some largely historical and now irrelevant
reasons for this, but it's not a trivial endeavour.

Also, I think there is a warranted apprehension among the desktop
search developers as to whether implementing a largely untested spec
is a good idea.  It's not really any different than the situation that
comes up whenever anyone wants to add API to GTK+.  That is, people
won't use the API until it's in GTK, but you don't want to add (and
freeze) an API in GTK if it's essentially unused and untested.  For
GTK, libegg has been a reasonable solution to this problem.  For
Beagle, that was the idea behind first implementing Xesam as an
adaptor rather than entirely ripping out our existing IPC mechanism.

Part of the problem is that we need more manpower.  We need someone
(or more than one person) to implement a Xesam-based client
application that is not designed with any one search system in mind.
Porting the existing abstracted backends in GTK+ and Nautilus would
also be a worthwhile task.  Only with those clients in combination
with the various search services that implement Xesam will we really
know how well it works in practice.

So get to work. ;) [1]

Joe

[1] FOCUS you BABOON! [2]
[2] http://mail.gnome.org/archives/gnome-list/1998-November/msg00294.html