Causes of session management problems in Plasma 5

[Andreas Hartmetz]

On Donnerstag, 26. November 2015 08:55:19 CET Thomas Lübking wrote:
> On Donnerstag, 26. November 2015 05:19:34 CEST, Nicolás Alvarez wrote:
> > What do you mean with "konsole asks"? Things like "You have
> > multiple tabs open, are you sure you want to quit?" and "You
> > have unsaved changes"?
> 
> Yes.
> 
> > If so, the scenario you describe is bad regardless of session
> > restoration.
> 
> Yes. Unfortunaltely.
> 
> > If konsole has to ask the user and the user has a
> > chance to say no and cancel the logout at that point, then
> > kwrite shouldn't have exited yet!
> 
> That's what the spec says, but the ksmserver change suggestion is
> about buggy clients that behave exactly this way.
> > If "app canceling logout" is a thing, then logout should feel
> > transactional to the user. Either logout happens and all apps
> > exit, or logout doesn't happen and nothing exits.
> 
> Yes. That's what the spec says ...
> 
> > I guess the implementation of that would be: all apps should be
> > given a chance to ask their questions and approve or stop the
> > logout
> That'S exactly what happens, but ...
> 
> > before *any* app exits.
> 
> We cannot stop process from doing stupid things, like exiting in the
> wrong moment.
> > This is how MS Windows works, by the way (or used to work).
> 
> No, it's how not buggy applications work on windows.
> 
> Cheers,
> Thomas

I was going to write pretty much the same things, Thomas was just faster 
:)
XSMP is a fairly well thought out spec although the full text, which I 
haven't read in its entirety, seems to be too long for what the spec is 
supposed to achieve. Logout being transactional is a big item in it.