Re: XFree86 5.0 TODO

On Thu, Mar 20, 2003 at 07:39:01PM +0100, Egbert Eich wrote:
> Sven Luther writes:
>  > :(((
> ???
> 
>  > 
>  > Would you care to explain a bit more about how this would work ? And in
> 
> I have no clear draft yet. So you'd have to wait.

Ok, no problem.

>  > particular why there cannot be a common set of functionalities, where
>  > chips not supporting them just not implementing them or something such ? 
> 
> Sure, there will be a registry where you can register attributes
> your driver offers. If another driver wants to offer the same it
> should reuse this attribute.
> You can query which attributes are offered by a specific driver
> and you can query the type and range of its arguments.

I personally think it is overkill, a simple tool which would do a
standard set of things would be simpler, with newer stuff implemented in
newer versions of the tool, and driver not supporting parts of it simply
ignoring it. But then, i agree this may not be the X way.

For example, i did originally write the glint/pm3 driver to use the
alpha channel for masking, but no app used it, and finally Alan rewrote
it so that a color key is used, altough you loose one color, thing which
would not have happened when using the alpha channel. You may well
export as many attributes as you like, the apps need to be handling them
all the same, so you have to upgrade the apps to take advantage of it.

What is really needed is a good set of standard attributes, or something
such.

>  > Or are there really all that much difference between how a matrox and a
>  > ati dual head card do it ?
> 
> Possibly not. But I would not rule out that you want to add something
> specific for a certain card where you can be pretty sure no other card
> will ever offer it.

I think it is more a generational thing, any new thing that gets added
in by one of the graphic companies will soon be copied by the others.
Some years ago matrox did stand apart with dual head, but now everybody
provides the same functionality, with different ramdacs and such.

Like explained in my mail exchange with Alan, i would like to see
complete independant viewport support, possibly with scaling. To have
that, you would need to define a framebuffer layout, possibly using
RandR to do depth switching and such, and then have a way to attach
viewports to them, dynamically. That is, you first attach one
viewport/monitor/head to the framebuffer, and can then choose to
activate the second head, and have it replicate the whole (or part of)
the display for doing a presentation on a video projector for example.
You can also have one viewport being a zoomed view of a part of the
screen for accesibility reasons and such. I believe such capacity is in
all of todays graphic card (it certainly did already exist in the G400).

Finally, we could do what Xv really does, it exports a certain number of
fixed functionality and adds support for additional attributes that can
be queried and such.

But again, i don't see what much can be added in this area, apart from
more precise zooms, higher frequency ramdacs, and more additional heads
(like parhelia's 3 headed solution).

Friendly,

Sven Luther