Re: Proposal for ConfigureRequest handling

[Dominik Vogt <dominik vogt gmx de>]

On Thu, Aug 08, 2002 at 12:35:56PM -0400, Owen Taylor wrote:
> 
> Hmm, so I guess there are two problems:
> 
>  - If DTWM and MWM do the wrong thing, it's probably hard
>    to get Java to just always do the right thing.

Actually, Java knows how to detect DTWM and MWM and switches to
the correct mode of configuring its windows.  We (the fvwm
developers) have tried for years to find out how Java exactly does
this, but never found a reliable way to make Java think its
running under mwm.
 
>  - It's probably hard to get window managers authors
>    to just let broken applications be broken.

Exactly.  That just won't happen with commercial WMs, and some
other WMs like fvwm that try to be compatible with everything.

> I really think it's best to say:
> 
>  - There is a single correct behavior for window managers
>    (what the ICCCM requires)
> 
>  - If this required behavior doesn't meet the needs of
>    applications, then we extend the NET WM spec to allow
>    alternate behaviors. 
> 
>    (I believe the ICCCM behavior is perfectly OK for placement,
>    but there are some brokeness in the Java specification
>    for how window sizing works that it can't handle.)
> 
>  - Window managers MAY use alternate placing behaviors
>    for particular apps that they believe are non-conformant,
>    but they MUST NOT do so for any app that indicates
>    that it supports the NET WM spec. (Insert definition of 
>    what indicates support for the NET WM spec.)

Fine with me.  That requires the application to publish this bit
of information.

>  - Applications MAY try to accomodate particular window 
>    managers with incorrect placement algorithm, but they 
>    MUST NOT do so if the window manager supports the NET WM
>    spec.

Okay.  Here, the window manager is resposible of advertising the
way it works.

Overall, I think this is a very good proposal.  No application and
no window manager has to implement any additional features unless
it already recognizes the CR problem and wants to do something
about it.  I'm all for it. 

> The burden has to be on window managers that care about
> ICCCM non-conformant applications, not on conformant
> applications or on window managers that only handle
> properly behaving applications.

That leaves an unspecified area:  what should the window manager
assume if the appication doesn't support the NETWM spec?  Im my
years as a WM programmer I noticed only one app that honours this
aspect of the ICCCM and does not support the NETWM spec (Java).
For all other (non-NETWM compliant), the non-ICCCM way was the
proper choice (of course I've only seen a small fraction of the
available apps).  I'd say it's up to the window manager to make
the 'right' choice (ignore the problem/hard code a table of
application features/make it configurable in the wm).

> There should no method of indicating "I'm non-conformant"
> because that's just silly.  If an app or window manager
> could be modified to indicate that it was non-conformant,
> it could just as easily be fixed to be conformant.

Good point.

Bye

Dominik ^_^  ^_^

 --
Dominik Vogt, mail: dominik vogt schlund de, phone: 0721/91374-382
Schlund + Partner AG, Erbprinzenstr. 4-12, D-76133 Karlsruhe