Path: utzoo!news-server.csri.toronto.edu!cs.utexas.edu!swrinde!zaphod.mps.ohio-state.edu!magnus.ircc.ohio-state.edu!csn!datran2!smb
From: smb@data.com (Steven M. Boker)
Newsgroups: comp.sys.next
Subject: Re: Compositing question
Message-ID: <1991Mar9.212912.3709@data.com>
Date: 9 Mar 91 21:29:12 GMT
References: <1991Mar9.000952.903@data.com> <Ibq3uOm00W02I6_LZK@andrew.cmu.edu>
Distribution: na
Organization: Data Transforms, Inc.
Lines: 41

In article <Ibq3uOm00W02I6_LZK@andrew.cmu.edu> wb1j+@andrew.cmu.edu (William M. Bumgarner) writes:
>If you dissolve by 50% each time, it will never reach zero, as you said.
> To actually dissolve completely, I suspect you need to have an ever
>increasing delta value as you do w/dissolve:toPoint:.
>

Here is what I have run into.  There is a difference in compositing with
2.0 and compositing with 1.0.  The simplest way to view this for yourself
is to compile CompositeLab on a 1.0 machine and also on a 2.0 machine.

Set the color of the source to be white and the transparency of the
source to be 20% (or so).  Now set the composite method to be SOVER and
set the transparency of the destination to fully opaque.  If you
now vary the color of the destination slowly from black to white you
will see the anomaly I am up against.

In 1.0 everything works as you would expect.  The portion of the destination
that is covered by the source is _always_ a lighter shade of gray than
the uncovered destination, until the entire rectangle has become white.

In 2.0 we see something different.  The portion of the destination that
is covered by the source is lighter than the portion of the destination
that is uncovered until the destination's color is about two thirds of
the way toward white.  After that, the source has _no_effect_ on the
destination.  In fact, you will see that the value of this lower limit
is dependent on the transparency value of the source.

My view is that this is a bug.  The metaphor of optical compositing can
no longer be used to understand the behavior of the compositing operations.
I will have to do a computationally expensive workaround which will limit
the complexity of smooth animation that I will be able to accomplish or
will require preprocessing of the animation as opposed to being able to
have the user interactively vary the transparency controls.

Steve.

-- 
 #====#====#====#====#====#====#====#====#====#====#====#====#====#====#====#
 #  Steve Boker           #                 En Vino Kaos                    #
 #  smb@data.com          #                En Kaos Veritas                  #
 #====#====#====#====#====#====#====#====#====#====#====#====#====#====#====#