kded won't start anymore

Sven Burmeister sven.burmeister at gmx.net
Sat Jun 21 07:57:37 BST 2008


On Friday 20 June 2008 23:38:54 Matt Rogers wrote:
> On Fri, Jun 20, 2008 at 10:17:36PM +0200, Sven Burmeister wrote:
> > One could create a plasmoid that checks every x seconds whether kded is
> > still running and displays a button to restart, if not. This way it would
> > use 0 space if kded is active and help the users that do not know what's
> > wrong if cookies, global shortcuts etc. do not work anymore.
> >
> > Sven
>
> Or kded could just restart on its own like plasma does. No point in
> exposing that to the user at all.

Actually I do not think this is a good idea, because it can fail and even 
worse it fails intransparently. The most robust solution is the best IMO.

Plasma only restarts twice within two minutes. So if a user adds a plasmoid 
which crashes plasma and tries again, because humans tend to try again, he 
needs to be lucky in order to not add it a third time or hit another crashing 
plasmoid. Otherwise, there is no plasma and if the user does not know, that 
what he has to start is called "plasma" or how to start it without a desktop, 
he is screwed, because you did not expose the actual issue to the user and did 
not offer a robust solution either.

The above restriction to auto-restart is necessary in order not to end up in 
an endless loop, which can happen too, if you just restart on your own without 
any restrictions. Imagine kded would crash at kde-start and restart itself 
endlessly.

So instead of hiding the problem, using a plasmoid or any other kind of 
information, would enable the user to understand that issue to the extent that 
he can avoid and report it, i.e. kded is crashing after I do xy. One could 
display a short text which explains that without kded neither cookies nor 
global shortcuts work and allow the user to restart kded, even though he might 
not know anything about  ALT+F2 or konsole. AFAIK kded can not be started from 
the menu, even if one would assume that a user knows the name of the app, 
which I consider highly unlikely.

Without this the user is put in danger of either ending up with no kded, an 
endless loop or not understanding the issue and hence reporting: my cookies 
don't work, global shortcuts stopped working instead of: doing xy crashes 
kded.

I hope this explains why I am in favour of transparency and mediating issues 
instead of implementing some magic which is not robust.

Sven




More information about the kde-core-devel mailing list