NSmooth Script - Searching for Testers and Feedback

As i transitioned to Blender 2.5 i found it hard that the NormalSmooth-Script was no longer available. So i searched for a solution and came up with my own script. It still has some minor issues (results are not perfect), but together with relax it can already do the job very well.

An Example

The testcase is a simple sphere where i dragged some vertices inward.
http://www6.pic-upload.de/thumb/20.01.11/a6w6xd1uq4i2.png

The script can be called with the “W”-Menu in editmode and works on the selected vertices.
http://www6.pic-upload.de/thumb/20.01.11/5m8vxbb4qx77.png

It is possible to customize the number of iterations and its strength. (Sliders have some issues. Don’t know if its my fault or Blender’s. Feedback is welcome. In any case u can enter the values directly.)
http://www6.pic-upload.de/thumb/20.01.11/7z4sp8k46b1s.png

As you might have seen, the result does not look very satisfying. But if you look at it from the side, you will see that it is close to what i want to achive and will fix later on.
http://www6.pic-upload.de/thumb/20.01.11/1p3wvtbflnmc.png

This is where the relax Script (mostly bundled with Blender already) comes in. Pleasing result?
http://www6.pic-upload.de/thumb/20.01.11/n4vhof3starz.png

Feedback
Any kind of feedback is welcome. Also take a look at the code, since im very new to python and think that i made some things far more complicated as they need to be.

Download

http://www.flickr.com/photos/91516124@N00/5374896048/Nice work. While some development is still needed of course, I think this tool is already useful as it is.

http://farm6.static.flickr.com/5201/5374896048_63e82692bc_m.jpghttp://www.flickr.com/photos/91516124@N00/5374896048/

I tried your tool on the side of a subdivided cube. and as I smooth it, the flat side is approximated to a sphere.

Yes i found something similar yesterday. Now i will have to find out how i could avoid situations like this. Also the performance could be better. Guess i have to ask some questions (will do it separately from this thread).