User bus conclusion

[Havoc Pennington]

Hi,

On Sat, Nov 13, 2010 at 9:40 AM, Matthew Johnson <dbus at matthew.ath.cx> wrote:
> On Wed Nov 10 17:05, Lennart Poettering wrote:
>> User logs in once. User logs in a second time. Both graphically. Now he
>> adds two applets to his first login. One RSS feed applet show his
>> favourite slashdot.org feed, the other one another RSS feed
>> spiegel.de. Now he logs out from the first login. Logs out from the
>> second. Logs in again. Should both applets now be shown? Logs in again
>> on the second as well. Should he now have 4 applets in total? He kills
>> an applet on one screen. Should the applet be removed on the other too?
>> Or only when he logs out? Or never? If he moves the applet around on one
>> screen, should it move around on the other too? Or should it only change
>> position after logout/relogin? Or would you keep it all time seperate?
>
> I don't disagree that there are issues here, but you know what, I don't really
> care what happens in this situation. Don't Do That Then. Pick a sensible and
> consistent option, then just keep that behavior. Maybe the first one wins, or
> the last one. Maybe you're prompted if they are different. Maybe you inotify
> the config so that when the first one is changed, it automagically changes on
> the second. There are clearly ways to solve this that aren't just "oh, it's too
> hard, lets just prevent you from doing that".

I think both of you have stated the argument well here. It's "should
apps/desktops (in theory at least) do something complex to handle
network homedirs" or "should apps/desktops say network homedirs are
too hard and just avoid a bunch of hypothetical code that probably
doesn't ever work anyway"

The reason I like the per-user session proposal, rather than the
traditional user bus proposal, is that the traditional user bus
_keeps_ the multiple session problem while _not solving_ the network
homedir problem.

Network homedirs are probably the only _actual_ useful reason to mess
with multiple sessions; multiple sessions, especially graphical
sessions, on the same machine are very marginally useful, at best.

I see the virtues of punting the multiple session problem _entirely_
and deleting a bunch of code. (Per-user sessions do this if you
declare network homedir off limits.)

I see the virtues of trying to handle multiple sessions because
network homedirs can be useful.

I don't see the virtue of solving multiple sessions on a single
machine, but leaving network homedirs unsolved. (This is what user bus
does.)

I simply don't see an approach that is best of both worlds. As best I
can tell there's flat-out a tradeoff here.

Havoc