feConvolveMatrix

What's the status of feConvolveMatrix?  Is there an ISSUE at the moment
regarding this filter effect?  Peter Graffagnino made a few suggestions
a while ago to fix problems related to kernels where the sum is zero.
There is another matter that I don't think anyone brought up.  The spec
at the moment in my opinion is not clear enough on how the convolution
should handle transparent pixels.  A pixel with an alpha of zero should
not contribute at all in the calculations.

Thank you.

Benoit Bezaire
benoitb@corel.com