Around 10 o'clock on May 12, [EMAIL PROTECTED] wrote:

> Isn't it less misleading to say partial coverage is indistinguishable
> from a transmissivity less than one.

Or partial coverage is indistinguishable from partial opacity, which is 
more-or-less what alpha measures.

> Translucency usually implies some degree of dispersive effects.  E.g.
> translucent objects cause blurring of the background object, not merely
> reduced light transmission.

That would be fun to implement, but it is a global effect rather than a 
purely local effect.  As we don't (currently) expect hardware to offer 
anything like this, there's no real reason to put it into the X server.

XIE provides some of this functionality, but the lack of acceleration made 
that extension worthless -- the extension could not ever gain traction 
because applications needed to provide a client-side implementation for 
servers that didn't offer the extension, and the extension offered no 
performance benefits.  Hence, the extra work needed to use the extension 
made no sense.

> The OpenGL specification states that alpha shall implement transmissivity,
> but the DirectX specification states that alpha implements voltage
> interpolation.  Since voltage interpolation is immensely easier to
> implement in hardware, all of the affordable PC graphics cards only support
> voltage interpolation.

Apparantly some PC graphics cards have started providing gamma-corrected 
compositing, which provides (slightly) improved appearance, especially for 
partial-pixel coverage approximations.

AA text suffers badly under voltage interpolation; I've already planned on 
adding gamma correction to the Render extension to improve the appearance 
of light text over a dark background.

> If we wanted to put in the effort, Render could define a software
> implementation (with some new pixrect definitions required) that truly
> did implement transmissivity computations.

Anything we can reasonably expect some common hardware to implement is 
certainly a candidate for inclusion in the extension, other operations
only move computation from client to server offering little in the way of 
significant performance improvements.

But, it certainly would be fun to design such a system and implement it on 
the client side as a part of a more comprehensive visually modelled 
graphics system.

Keith Packard        XFree86 Core Team        HP Cambridge Research Lab


_______________________________________________
Render mailing list
[EMAIL PROTECTED]
http://XFree86.Org/mailman/listinfo/render

Reply via email to