Re: [wmii] Re: layout per tag.

On Tue, 08 Aug 2006 09:30:52 +0200
"Stefan Tibus" <sjti_AT_gmx.net> wrote:

> On Sat, 5 Aug 2006 09:28:49 +0200 Uriel wrote:
> > There are plans for a feature similar to Dump in acme which is
> > probably what you want.
> > On 8/5/06, Stephan Maka <stephan_AT_spaceboyz.net> wrote:
> > > Anselm R. Garbe wrote:
> > > > [...] For
> > > > session info use swuspend or suspend2 in recent Linux kernels.
> > > > There is no need to reboot the whole computer nowadays, even if
> > > > you switch it off.
> > >
> > > But when software crashes just like wmii-3.1 does from time to
> > > time here on a FreeBSD machine (haven't gotten into debugging as
> > > this seems not reproducible enough) session saving becomes useful.
>
> Another interesting use of session saving and tag-associated
> layouts is on multi-user machines. On such a system I will have
> to log-off sometime. And after log-on I might either want to see
> the programs I had open in my last session or just have the same
> layouts predefined without the programs open yet. (I'd prefer
> the last one, as it makes log-on faster and I'll start programs
> when I need them only - nowadays software is too large to have
> many of them open in the background just waiting to be used.)
> I have to admit this goes into a direction similar to wmi, but
> in fact there are tasks, which stay the same for long times and
> there fixed layouts could be very useful (while the less static
> tasks should be handled dynamically, of course).
>

I have to disagree completely. There should be no fixed layouts in the
window manager, in any case. There were reasons this was dropped
in earlier wmii versions. Rather, the applications should provide for
enough hints (and the wm should listen to them, of course, while
placing them into their views) to dynamically create usable layouts.
Fixed layouts are a totally wrong approach to this problem.

In addition, there are just too many application states a session
saving algorithm of a window manager just can't take care of. This would
lead to inconsistent and confusing behaviour, being more often an
obstruction than a help (seen this often enough on my
brother's KDE box.)
Suspending programs should thus work on some other level, maybe the
application or the OS level in RAM. [1] discusses some methods to
achieve this, though it a) isn't really that magic thingie we're
looking for and b) won't solve the window management problem anyway.

Greetings
Denis

[1] http://www.cis.ksu.edu/~dan/dump.pdf