From: Jon M. Taylor <taylorj@ecs.csus.edu>
  To  : Tristan Wibberley <twibberley@llamacom.com>
  Date: Sun, 30 Aug 1998 14:56:01 -0700 (PDT)

Re: New function in libggiMisc RFC.

On Sun, 30 Aug 1998, Tristan Wibberley wrote:

> Marcus Sundberg wrote:
> > 
> > Tristan Wibberley wrote:
> > >
> > > Hi all,
> > >
> > > I think something is needed for greater window system support (in X and
> > > eventually berlin).
> > >
> > > It would be nice to see sets of widgets that use ggi to draw. I propose
> > > one new call in ggiMisc and some new struct definitions.
> > >
> > > The call would do nothing on most targets, but on X you fill in a struct
> > > with some data such as title, geometry, borderstyle and pass that to
> > > this function which the target implements. GGI should not specify the
> > > behaviour but the target should - ie, you need to know what you're
> > > doing, but it's not complicated.
> > >
> > > eg: (pseudocode):
> > >
> > > #define GGI_WINSTD structsize;
> > >
> > > typedef void *ggimisc_win_t;
> > >
> > > typedef struct {
> > >         GGI_WINSTD;
> > >         title;
> > >         resizeable;     /* not by default no way, but allows
> > >                            you to resize and check mode on EvExpose */
> > >         geometry;
> > >         winstyleflags;
> > > } ggimisc_xwin;
> > >
> > > typedef struct {
> > >         GGI_WINSTD;
> > >         title;
> > >         titleface;
> > >         blah;
> > > } ggimisc_berlin;
> > >
> > > void setupwin(gg_visual_t vis)
> > > {
> > >         ggimisc_win_t windata;
> > >         ggimisc_xwin xwindata;
> > >
> > >         /* fill the xwindata struct */
> > >         windata = (ggimisc_win_t)(&xwindata);
> > >         ggiSetWinData(windata); /* copies the data ready for use if mode not
> > >                                    set yet, else uses the data */
> > > }
> > >
> > > We need a way to tell what target ggiOpen(NULL) has chosen, is that
> > > possible already. A programmer needs to be able to choose how to fill in
> > > the struct for the windowsystem being used.
> > >
> > > This definately needs a lot of thought, but I think we should implement
> > > something like this.
> > 
> > Yes, I've been thinking of window-titles and such before.
> > Putting sucha function in ggimisc is a good idea, but the
> > structure passed should _NOT_ be targetdependent. This defeats
> > the whole purpose of GGI. Parameters that doesn't make sense to
> > the particular target will simply be ignored, and their entries
> > will be set to NULL/some other invalid value uppon return.
> 
> Agreed.
> 
> > So what does this structure need?
> > 
> > typedef struct {
> >         char      *title;
> >         char      *iconname;
> >         ggi_coord  position;
> >         char       resizeable;
> >         char       borderless
> 
>           void      *targetdata;
> 
> > } ggi_wininfo;
> 
> We at least need to allow for a target to accept non-standard data (we
> don't know what might be provided by other targets 

	Words to live by.  Implement a generic medatada-passing mechanism 
and you don't have to worry about the data structure.

> such as berlin) -

	Berlin is not a target.  It is a LibGGI-using application.

> just let that be not-guaranteed. The first field of the target data
> should be guaranteed to be an int containing an identifier to indicate
> which target this data is for - so if you provided data expecting X,
> berlin knows to ignore it. Perhaps the second field of targetdata should
> be guaranteed to point to another targetdata struct or void - Now a
> programmer can load data for all the targets he wants to support.

	Yes.  Exactly.  A metadata type identifier and and a void pointer
to the metadata struct.  All you need. 

> ps, did this message have twibberley@llamacom.com as the reply-to
> instead of ggi-develop@eskimo.com? (I have CC'd this there).

	Yes.

Jon

---
'Cloning and the reprogramming of DNA is the first serious step in 
becoming one with God.'
	- Scientist G. Richard Seed