Lars Hellström <[log in to unmask]> writes:

>   \@fancyfor\@tempa:=<list>\do{\@tempa}{, \@tempa}{ and \@tempa}
> and for <list> equal to A; A,B; A,B,C; A,B,C,D respectively get the results
>   A
>   A and B
>   A, B and C
>   A, B, C and D

As it happens I was working on similar functionality recently for the
amsrefs package (ftp://ftp.ams.org/pub/tex/amsrefs.zip). The package has
a \PrintSeries function used to print author names in a bibliography
when the data is a list of names in the form

  \name{Downes, Michael}\name{Hellström, Lars}...

For AMS conventions your first/middle/last approach is not quite
adequate because cases 2 and 3 must be printed as

  A and B
  A, B, and C         (Note the extra comma.)

And there are some other complications, as explained in the following
commentary from amsrefs.dtx:

%    The \cn{PrintSeries} command prints a list of objects in series
%    form. The essential idea is to produce something like \qq{A, B, and
%    C} when we are given three elements A B C, with suitable variations
%    in the punctuation and other intervening material if the number of
%    elements is less or more than three.
%
%    To generalize this process, we envision \cn{PrintSeries} being
%    called as

%\PrintSeries{i0}{i1}{i2}{i3}{i4}{\do{A}\do{B}...}

%    where i0, \dots, i4 are material to be interpolated and the last
%    arg is a list of indeterminate length where each element consists
%    of a macro and its argument. The output, depending on the number of
%    elements, will be

% i0 A i4                 % 1 element
% i0 A i1 B i4            % 2 elements
% i0 A i2 B i3 C i4       % 3 elements
% i0 A i2 B i2 C i2 ... X i2 Y i3 Z i4   % 26 elements

%    That is the simple explanation but in practice there are some
%    additional complications. What if user-supplied line breaks have to
%    be supported at the boundaries between elements? What if in
%    addition to adding material between elements we also want to apply
%    some handy function to each element (e.g., \cn{textsc})? Even
%    worse, what if we want the function to be different depending on
%    the position of the element in the list?? Indeed if this did not
%    happen to be the case with the current application I would not have
%    gone to the extra trouble of supporting it. But if it must be so,
%    then the output that we need from a list \verb'\do{A}\do{B}...' is

% f0{A}
% f0{A} p1 i1 f1{B}
% f0{A} p2 i2 f2{B} p3 i3 f3{B}

%    and so on, where
% \begin{itemize}
% \item $f_n$ is a macro taking one argument,
% \item $p_n$ is punctuation\mdash material that must precede a line
%    break if one occurs at this boundary,
% \item $i_n$ other interpolated material, as before.
% \end{itemize}
%    To reduce the number of distinct required objects we decree that
%    each element will get braces wrapped around it as a matter of
%    course; then it is possible for f1, f2, f3 to be assimilated into
%    the tail end of i1, i2, i3.