> This is to point at a shortcoming of NFSS2: (I don't know whether
> this has already been discussed when designed NFSS)

There are many potential axes that NFSS could have incorporated.  I
don't remember enough of the original papers to be sure off hand how
the present set were arrived at.

> There is such a thing as a slanted/italic small caps font
> (see Sebastian Carter's "Typographers on Type" for an example:
> it contains a page showing different fonts of the Zapf Renaissance (1986)
> family, including such a font).

So you reckon small caps is "just another axis"?  Well, you're not
alone...

> So it doesn't seem appropriate to classify small caps as a shape,
> rather, I think, it should be considered as a case. Thus resulting
> in control sequences like
>   \textcaps{...} and {\capscase ...}
> or the like, analogous to uppercase and lowercase.

Except that uppercase and lowercase are TeX primitive mechanisms with
truly eccentric properties, which can't be used in a modal fashion, as
in your second example.

> With the current fonts and current NFSS, it is e.g. not possible
> to markup names with small caps ... [with] italics

Quite so.  However, adding an axis to NFSS to take account of *one*
oblique SC font seems to me excessive.  I don't believe that NFSS is
perfect, but I don't think it's bad as it is.

> Unfortunaly I can't think of any way to incorporate this easily
> in NFSS and especially in its current interface.

An N^{3}FSS (for LaTeX 3) could perhaps consider the possibility of
variable numbers of axes for font families.  To incorporate it in
N^{2}FSS (the one in 2e) would imply changing pretty much all the code
to add a fixed extra axis to every font.  It would be worse than the
font changes between 2.09 and 2e.

Robin Fairbairns