I tend to agree with this: if the user binds a new name to the transformed object, then he would expect that the new object can be changed independenly. The design of the pyFormex language was purposely done in that way for the geometry of the objects. So we should do the same for the attributes.

This will however mean a lot of copying of attributes. Think about a long chain of transformations with a copy on each step:

G = G.trf1().trf2().trf3().trf4().trf5()...

But the the same holds for the coords, and the attribs will usually be limited.
And, the user can set the attributes after most of the transformations.

So I will go ahead and let the attributes be copied.
I think for now we will only implement it in simple transforms that just do some coordinate transformations.

With the .attrib likely to become more popular in pyFormex, and the draw() arguments only being a temporary override of the attributes set on some Geometry, this question becomes urging again.
Suppose a color is set on a Geometry:
G = Geometry().attrib(color=red)
and the Geometry is the drawn in a scaled version:
draw(G.scale(0.5))
The color will not be red, but the default (black).
This is likely not what the user expects.

This also means e.g. that the customization of the Triade can not be colored (except by the way of prop color), because the Triade geometry is always scaled in the rendering.

So I think inheritance of .attrib should be implemented.

But should we make a copy, or just use the original set of attributes?