Re: Proposal for the geometric engine, RFC

[Dylan Paul Thurston <dpt@math.berkeley.edu>]

On Wed, Mar 15, 2000 at 09:35:30AM -0800, Dylan Paul Thurston wrote:
> On further investigation, this seems to be only a drawing bug.  Even
> when the points are exactly collinear (to screen resolution),
> intersections with other lines are drawn (nearly) correctly, existing
> and not existing when appropriate.

To clarify what I mean be "nearly": if the three points are all on a
horizontal line, the arc itself does not appear, and intersections
with another segment sometimes appear one pixel too low.  I suspect
this is a rounding issue: maybe you're slightly wishy-washy about
whether coordinates refer to centers of pixels or their lower left
corner or some such.

Oddly, Dr Geo refuses to create an arc through three points that are
constrained to be collinear, even in complicated ways (like Desargues'
Theorem.)

Best,
	Dylan Thurston