Re: Gnome is not easy to use in 800x600

[textshell t-online de]

"Brandon S. Allbery KF8NH" wrote:
> 
> On 09/05/00 12:03:01 +0100, John Harper <john dcs warwick ac uk> wrote:
> +-----
> |
> | Frank Felfe writes:
> | |It is my understanding that the WM manages run time window data. I
> | |still think, that persistent window data must be stored by someone
> | |else, otherwise the WM would take over session management tasks. I
> | |thought the point of sawmill is to manage runtime window data only and
> | |leave session management to other programs, but I might be wrong. Is
> | |this kinda stuff specifies somewhere as standart?
> |
> | The ICCCM dictates that the wm should save session data (window layout,
> | stacking order, etc..)
> +--->8
> 
> The ICCCM is an ivory-tower disaster that proves that decreeing behavior
> without actually testing it in the real world doesn't work.  The vast
> majority of X11 app misbehaviors I see are because someone decided that
> they should slavishly follow the ICCCM and/or the XSM spec even when they
> are clearly broken.
> 
> In particular, the specification that the WM maintain window positions has
> so many flaws it is ludicrous.  Some examples:
> 
> - you can't change window managers without losing your saved window
>   information
> 
> - wm/sm interaction is undefined and problematic
> 
> - interaction between both wm/sm and USPosition/PPosition is apparently
>   sufficiently problematic that user-specified geometry doesn't work
>   reliably in at least some apps once a window has been positioned manually
>   (reproduced with gnome-terminal and sawfish; I suggested the person who
>   tripped over this mail to gnome-devel-list, but he hasn't yet that I've
>   seen).
> 
> Granted that we have no good alternative to the ICCCM or the XSM spec, but
> following either when they are obviously wrong can only be called stupid.
> 
> (This is a general X11 rant and is not limited to GNOME, by the way; there
> are similar examples of the insanity imposed by slavish ICCM/XSM compliance
> in KDE, CDE, etc.)
> 

Well I think the right approach is to have some shared responsibility. I
think the sane way would be to have some way an application can request
the windowmanager to record it's windows states. The Application should
store this and pass it back to the WM when appropriate maybe with some
application specific settings (e.g. User explicitly state use this). The
WM would than react according to it's policy. This way the Appliction
can specify the windows where this makes sense, and just skip those
where this is just stupid(e.g. messageboxes with changing content
length).

This is not meant to replace session management, but this fills the gap
that quiting an application and later restarting it leaves.

It would even be possible to have an standard for this data that all
GNOME WMs would follow so that the basic settings (pos, size, sticky)
would persist even when using another WM, while the WM can save all the
associated state of an window thus making it much easier to have an
consitetent behaviour. 


Just some thoughts for some one who does not really know much about WM
programming, hope this gives sense...

   Martin