[PATCH] dbus-glib patches

David Zeuthen david at fubar.dk
Mon Mar 17 10:21:01 PDT 2008

On Mon, 2008-03-17 at 12:20 +0100, Rob Taylor wrote:
> > The second one is needed if you want to use the generated client glue
> > and you have operations that lasts longer than the default timeout.
> Hmm, this isn't as useful as it seems, as (iirc) the bus daemon also has 
> a 25 sec timeout on method calls. Those kind of operations are better 
> structured as a call to start and a signal to stop. Also long operations 
>   often need a cancel call as well...

It really depends. Some operations 

 - can't be sensibly be canceled (e.g. suspend, hibernate).

 - are highly destructive (e.g. partitioning resize) so it's not
   feasible to allow canceling unless you want performance to
   suck (e.g. backup all data elsewhere before the op)

 - are sensitive (password, others) and emitting a signal would
   create a covert channel (though Colin mentions on IRC having
   a signal addressed to the sender only could work)

This pattern also increases complexity in both the server and the client
and it also breaks dumb clients such as shell scripts using dbus-send


More information about the dbus mailing list