[Gs-devel] x11alpha device

Raph Levien raph at levien.com
Fri Oct 12 09:26:24 PDT 2001


On Fri, Oct 12, 2001 at 09:12:56AM -0700, L. Peter Deutsch wrote:
> The x11alpha device is **NOT** equivalent to the x11 device with
> TextAlphaBits or GraphicsAlphaBits set.  x11alpha adds an alpha value to its
> native pixel representation, and is supposed to (but does not) actually
> store an alpha plane to parallel the screen buffer.  I.e., it supports
> map_rgb_alpha_color and setting an alpha value in the graphics state.
> 
> The NeXT/DPS alpha and compositing machinery is not in very good shape, and
> it may overlap with the PDF 1.4 transparency code, but I don't think it
> should be abandoned piecemeal.

Now I'm confused.

As I see it, the x11alpha device is a broken hack which does an ok job
displaying antialiased text, but tends to mangle vector graphics, in
some cases badly. By contrast, x11 with alpha bits set does an equally
good job with aa text, and also renders graphics well.

Unless I'm misreading the code, x11alpha has nothing to do with the
NeXT/DPS alpha machinery. I'm unaware of any files which will actually
render better in x11alpha than x11 with the alpha parameters.

As I say, I'm puzzled why you seem to be attached to this broken
hack. I believe the most good it's ever done over the years is to
mislead people into (quite reasonably) believing that Ghostscript has
no idea how to do alpha compositing correctly.

Also, in regards to Ray's comment: try knight.ps from the data/misc
directory if you don't believe that -dMaxBitmap=50000000 makes a
dramatic speed improvement, even in the local case.

Raph




More information about the gs-devel mailing list