Re: Circular paths

Dirk Schulze:
> If you want to have consistent behavior between viewers a minimum
> should be the number ob cubic Béziers (I think we all agree that
> cubic Béziers are the best solution for normalization, but even that
> is not specified). Depending on the arc, one to three cubic Béziers
> are necessary to normalize an arc (see older comments on the list).
> How many degree of the ellipse should be covered by every curve?
>
> We need some kind of algorithm in the specification. We already have
> an algorithm on WebKit that normalizes an arc for drawing it to a
> surface. While the behavior matches the behavior of Opera most of
> the time, we have examples where is doesn't (number of segements are
> different). And this will cause different normalized pathSegLists
> between viewers.

Yeah, I’d say that probably having the number of segments be consistent
across implementations is more important, given that could really trip
people up who are modifying the list.  Can it ever be the case that, for
a given number of Béziers, you get a more accurate approximation if you
don’t distribute the degrees evenly among them?

-- 
Cameron McCormack ≝ http://mcc.id.au/

Received on Tuesday, 4 January 2011 21:58:38 UTC