[Xr] Xr API comments

[Keith Packard]

Around 15 o'clock on May 27, Carl Worth wrote:

> If there were a formal Xr specification, it would guarantee that the
> effect of drawing with an arc would deviate at most "tolerance" device
> pixels from the geometrically correct result.

One additional useful constraint is that an arc from 0 to x trace a path
which is (really close to) a subset of the path traced by an arc from 0 to 
x + dx.  I don't think it matters whether we define arcs in terms of 
splines or lines; although splines won't involve any joins if done 
correctly.

> Hmmm... After typing that, I'm wondering if we want to come up with
> a definition for tolerance other than device pixels that would make
> more sense for ridiculously high-resolution devices.

No, I think ridiculously high-resolution devices should generate 
ridiculously high-precision output.  That's how postscript is defined.

> I think circular arcs are a special case that merit direct support.

Circular perhaps, but elliptical arcs have axes to define, which are a 
pain.

-keith