Re: [Tracker] status of 0.6.90?

[Martyn Russell <martyn imendio com>]

Tshepang Lekhonkhobe wrote:
> On Wed, Oct 8, 2008 at 4:27 AM, Saleem Abdulrasool
> <compnerd compnerd org> wrote:
> > The TrackerConfiguration API is far easier to maintain.  Rather than having
> > functions for every single option, its a simple, generic API.
>
> It's interesting that you bring this only now, months after
> tracker-config has been established as THE config API for Tracker. I
> prefer tracker-config 'cuz it allows API user write less code and also
> more flexible (IE if the details of the config files, say EnableWatch
> to EnableMonitoring, no user-side code needs changing). Please correct
> me if I'm wrong.

I just removed TrackerConfiguration and patched the source in favour of
TrackerConfig.

TrackerConfiguration is flexible enough to write any kind of data you
want, but that's not what we want and it is error prone because it means
every place that uses the config has to know what string a configuration
has to be able to get it and set it correctly.

There are other implications too. This API is exported to 3rd party
application developers and the TrackerConfiguration API gives 3rd party
application developers no idea as to what configuration they can
get/set. It also means that they could potentially add all sorts of
unwanted data into the Tracker .cfg file.

The only reason it is easier to maintain is because the burden is
shifted from the config API to the developers using it.

There are advantages to the TrackerConfig object too. If at some point
we so choose, we can export it to DBus and allow applications on the
desktop to get/set the configuration with just a few lines of code.

-- 
Regards,
Martyn