Please save your time and energy on this question, I also submitted one possible solution.
Zdeněk Wagner, a Czech TeXist, in his article (the Zpravodaj journal of the Czechoslovak TeX Users Group, 1/2013) mentions a problem of getting proper outline font, e.g. for Devanagari and Arabic fonts, by using the \pdfliteral command. This is an example and a preview typeset in the CODE2000.TTF font.
The problem is that we actually see how glyphs are put together, see picture on left, the correct form is shown in picture on right. How to get from the left picture to the right one? Why this example is running with lualatex, but not with xelatex?
%! {xe|lua}latex mal-deva.tex
% xelatex is not working, lualatex does work
\documentclass[a4paper]{article}
\pagestyle{empty}\parindent=0pt
\usepackage{fontspec}
\usepackage{xltxtra}
\begin{document}
\huge
% http://web.archive.org/web/20101122142710/http://code2000.net/CODE2000.ZIP
\setmainfont[ExternalLocation]{CODE2000.TTF}%
\def\malword{देवनागरी}% Term in English: Devanagari
\pdfliteral page {q 2 Tr 0.7 w 0 0 1 0 k 1 0 1 .1 K}
\malword
\pdfliteral page {Q}
\end{document}

! Undefined control sequence. l.10 \pdfliteral page {q 2 Tr 0.7 w 0 0 1 0 k 1 0 1 .1 K}\malword\pdfliteral{Q}. Is there something simple I am doing wrong? – Steven B. Segletes Apr 04 '14 at 23:35xelatexbut it works withlualatex. I am updating the question as well as my answer. – Malipivo Apr 05 '14 at 05:21