Re: Module proposal: dconf

[Bastien Nocera <hadess hadess net>]

On Wed, 2009-10-14 at 11:46 +0100, Emmanuele Bassi wrote:
> On Wed, 2009-10-14 at 11:24 +0200, Matěj Cepl wrote:
> > Dne 13.10.2009 22:42, Dan Winship napsal(a):
> > > OMG ITS TEH WINDOWS REGISTRY!!!1!1II|! IF ANY APP WRITES A SINGLE BYTE
> > > WRONG THEN ALL OF YOUR APPS WILL BREAK AND YOU WON'T BE ABLE TO LOG IN
> > > ANY MORE 
> > 
> > +1 :)
> > 
> > People who are not able to learn from history are doomed to live through
> > it again.
> 
> oh, puh-leaze. I think we have learned some things in the past 10 years.
> 
> if you bothered to look at GSettings and dconf you'd realize that there
> is only one process allowed to write to the database: no application
> will ever write to the database - only read operations go through the
> library, to avoid contention.

How can we be sure to have only one process writing to the database
when:
- using NFS home dirs and not sharing displays or machines
- using remote X apps over ssh

Those were problems with GConf, but at least it didn't have the
"everything will break" problem if the above wasn't true.