[metapost] MF interpolation instability

[Boguslaw Jackowski]

> Also I don't know if  it make sense, but at least it looks symmetric and
> it's correct.

Very many thanks, Luigi! Your examples are very useful as an argument 
that, in principle, polygonal pen envelopes can be calculated correctly,
at least the asymmetry reported by Stephan is not unavoidable.

There is a tough problem, however, with the endpoints, as this requires 
employing a "removing overlaps" procedure which, as the experience shows, 
was nowhere implemented robustly as yet.

Alternatively (and rather ugly), we could decide that additional paths, 
namely, pen countours, are being drawn at the endpoints.

Cheers -- Jacko

Ps. As to removing overlaps, a few years ago I suggested that the procedure 
finding a common outline could be a hybrid creature, i.e., it could 
operate on both outlines and a bitmap, using the bitmap as an auxiliary 
mean to decide which segments of the outline are relevant. I still believe 
that it might be a satisfactory solution, unfortunately, I haven't 
tackled the problem seriously yet, sorry...

-- 
BOP s. c.
ul. Bora-Komorowskiego 24, 80-377 Gdansk, Poland
tel./fax (+48 58) 553 46 59
bop at bop.com.pl, http://www.bop.com.pl