Re: [Evolution-hackers] e_cal_remove_object_with_mod() + empty rid: semantic?

[Matthew Barnes <mbarnes redhat com>]

On Fri, 2011-05-13 at 12:48 +0100, Ross Burton wrote: 
> On 12 May 2011 11:44, Patrick Ohly <patrick ohly gmx de> wrote:
> > It'll lead to a change of the D-Bus API. For "master" that shouldn't be
> > a problem, but I also want this in MeeGo for KCal-EDS, based on 2.32. I
> > guess we have to bite the bullet and maintain a MeeGo version with a
> > different API than regular 2.32.x.
> 
> I've always said that the DBus interface isn't a stable API and
> shouldn't be trusted to remain constant.  If nobody else documented
> that the DBus interface is stable then we shouldn't worry too much.
> As a DBus API designed to be used by anything other than
> libebook/libecal it's pretty dire.

It's fine to change the D-Bus API as needed, but remember to bump the
version number in the bus name.  I only recently added version numbers
after the whole 2.30 -> 2.32 upgrade debacle, so we're still at:

   org.gnome.evolution.dataserver.AddressBook0
   org.gnome.evolution.dataserver.Calendar0

The bus names are defined in evolution-data-server/configure.ac.

This avoids the problem of users upgrading a client package like
Evolution while the older D-Bus services continue running.  If the D-Bus
interface changed, now suddenly the client can't talk to the services
correctly anymore.  That's what caused all the problems with local://
ESource URIs in 2.32.