multiple local X logins by same user (was: Tried Pulse Audio Again--No Good For A11y)

Matthew Woehlke mw_triad at users.sourceforge.net
Wed Sep 24 00:38:09 UTC 2008


Lennart Poettering wrote:
> You did not respond to my question what you'd think the proper
> behaviour would be for gnome-panel. I'll take that as an
> acknowledgment that you understand that the problem exists.

I don't use gnome, but... I'd expect plasma (KDE's desktop container and 
taskbar) to write its settings out when I apply changes. This should 
either overwrite any other changes made to the file (a common behavior 
when running multiple independent instances of a program) or else merge 
the changes in a way that doesn't break the app the next time it starts 
up (but is not necessarily required to have any "desired effect"; IMO 
that's permissible for such a usage.)

Therefore, I don't consider that there is a significant "problem" here. 
Ideally, updating the configuration would be atomic (this is probably 
needed regardless, although you aren't likely to have race conditions 
when talking about user settings) and would notify all instances to 
re-read the configuration, but I don't consider this crucial.

>> Windows terminal services has gotten this more or less right since at least 
>> windows 2000 server that included 2 licenses for administrative use.  If 
>> they can do it with an interface that wasn't designed to be remote or 
>> multiuser, it can't be that hard.
> 
> Are you sure you can log in twice on Win2k as exactly the same user id?

I don't know about w2k, but it works in 2k3 (I just tested). Two logins, 
same user, different apps running. Wow, it's almost like having virtual 
desktops! :-)

So, er... yeah. To throw in my $0.02, I don't see why this shouldn't be 
permitted.

(Incidentally, Firefox drives me nuts not letting me run it on my local 
login and via 'ssh -X' from a remote machine. Some apps are even worse 
behaved than not coping with two full local X sessions as to go so far 
as to not permit multiple instances to be run /at all/.)

> On Tue, 23.09.08 15:50, Les Mikesell (lesmikesell at gmail.com) wrote:
>> But, if it can't be done right, the WM should enforce it and give you a 
>> choice of killing the old session when you attempt a new login instead of 
>> just letting random things fail.
> 
> Nah, if at all that's the job of the dm or the sm, not the wm.

Er... I guess. I suppose the DM has to figure out how to start a second 
X root session. Beyond that, apps that break when you try to run two 
instances in multiple X sessions are, IMO, just plain buggy. It's 
perfectly normal for me to run apps over remote X, and not uncommon to 
want to run Xephyr sessions. I expect apps (and this includes WM's and 
other desktop bits) to work just fine if I try to run multiple instances 
in multiple X sessions, regardless of what kind of X session I'm trying 
to run it in (local, nested, remote, headless*).

(* i.e. vncserver, or any others that use an offscreen buffer for X output)

-- 
Matthew
Igor Peshansky: Don't hippos love water even more than dogs?
Dave Korn: Don't ask me. I didn't even know that hippos loved dogs.




More information about the fedora-devel-list mailing list