small phi font problem

I am working on a courseware of calculus for university students,  
based on xml technique, including mathml.

In math the common curly small letter phi font is extensively used as  
the polar angle in some coordinate systems (for example -  
cylindrical), and another one - a miniaturized capital phi is used for  
different purposes. It is imperative that any browser, in any  
appropriate operating systems, uses the same rule for rendering these  
fonts. That way a user can interchange the browsers without affecting  
the rendering.

With the increasing browsers' support of mathml, and with the  
different interpretations of the two fonts of the small Greek letter  
phi, a problem arises. Details about the fonts and their  
interpretations by the mathml rendering software are shown in the next  
snapshot (a gif file):
"http://alzt.tau.ac.il/~dagan/mathanimated/phifont/phi.gif".
Any one can check how his own browser renders the small phi by applying
"http://alzt.tau.ac.il/~dagan/mathanimated/phifont/phiFont.xhtml".

The identical way of rendering the fonts was achieved so far by IE +  
MathPlayer, and  by the FireFox-browser version 2. For some obscure  
(to me) reasons the rendering of the phi fonts in a text and in a  
mathml expression are inverted there, but since the same rule is  
applied by the both, the interchangeability between the browsers is  
kept.

One could expect that the browsers with native mathml support in their  
beta stage of development -  FireFox version 3 and Opera version 9.5 -  
will continue to follow the same rule for rendering the phi fonts, in  
order to be interchangeable among themselves and with respect to the  
already existing ones. Unfortunately this is not the case at present,  
as seen in the above mentioned snapshot
"http://alzt.tau.ac.il/~dagan/mathanimated/phifont/phi.gif".
Even the same browser acts differently in different platforms (Mac and  
Windows).

If FireFox will return and readopt its own rule, used in FireFox  
version 2, it is conceivable that Opera will follow, and the problem  
will be solved. I avoided to file a bug-report on this subject, it  
seems to me that an email to this list is more appropriate.

Cheers, Samy