[MathML3-last-call] mathvariant

I really don't like the philosophy behind the paragraph on  
@mathvariant, specifically, the statement [1]:

    By design, the only cases that have an unambiguous interpretation
    are exactly the ones that correspond to SMP Math Alphanumeric Symbol
    characters, which are enumerated in Section 7.5 Mathematical  
Alphanumeric
    Symbols. The mathvariant values "initial", "tailed", "looped" and  
"stretched"
    are expected to apply only to Arabic characters. In all other  
cases, it is
    suggested that renderers ignore the value of the mathvariant  
attribute
    if it is present.

There are three ways a renderer can satisfy a request for a particular  
mathvariant.

1) The desired glyph may be found as a variant-glyph --- in some font  
that the renderer has available ---
at the same code-point.
2) The desired glyph may be found --- in some font that the renderer  
has available --- associated to another code-point (say, in Plane-1,  
as discussed by the current text).
3) If the desired glyph cannot be found, perhaps some transform  
(overstriking, say, in the case of bold) may achieve an approximation  
of the desired effect.

I don't think the Spec should be micromanaging how a renderer is  
supposed to deal with @mathvariant.

If the renderer has the desired glyph available to it, it should make  
use of it (whether through mechanism (1) or (2)). If it doesn't have  
the desired glyph available, "MathMLforCSS" is not magically going to  
solve the problem.

As a specific example, the current Spec says that

    <mi mathvariant="bold-italic">&imath;</mi>

should be ignored, despite the fact that

a) Many fonts (STIXGeneral, Verdana, Trebuchet, Times, Palatino, ...),  
have a bold-italic variant-glyph for U+0131.
b) This is a perfectly reasonable variant to request.

Jacques Distler

[1] http://www.w3.org/TR/MathML3/chapter3.html#presm.commatt