What is the implication of the warning in the fontspec manual with respect to babel and LuaTeX?

In the fontspec manual, there is a section on LuaTeX that states:

babelThe babel package is not really supported! Especially Vietnamese, Greek, and Hebrew at least might not work correctly, as far as I can tell. There’s a better chance with Cyrillic and Latin-based languages, however—fontspec ensures at least that fonts should load correctly, but hyphenation and other matters aren’t guaranteed. Under XeTeX, the polyglossia package is recommended instead as a modern replacement for babel.

What are the technical implications of this? Should I be afraid of anything?

-
What languages do you need? –  egreg Oct 6 '11 at 12:33
Currently I don't need any languages (except German), I just want to know what this means. What is broken? Why is it broken? I want to be able to give an answer if someone asks me: "Can I use fontspec with Vietnamese and if not, why not?" –  topskip Oct 6 '11 at 12:35
And you don't like the answer "Yes, but you should use polyglossia instead of babel"? –  Ant Oct 6 '11 at 12:47
@Ant no, because the why part is what I am interested in. :) –  topskip Oct 6 '11 at 12:49
@Ant: No, because polyglossia doesn't work (well) with LuaTeX. –  Martin Schröder Oct 7 '11 at 10:05

Languages that set an output encoding as part of their working in babel will have problems. For example, \selectlanguage{vietnamese} issues

\fontencoding{T5}\selectfont
\def\encodingdefault{T5}
\language\l@vietnamese


and of course the first two lines are nonsense with fontspec and OTF fonts. The same problem along with others will appear with languages written in other scripts (Cyrillic, Greek, Hebrew). See, for example, Table of contents encoding issue with LuaLaTeX.

-
Is this it? No more problems? Seems really easy to fix. –  Mateus Araújo Oct 6 '11 at 15:54
@Mateus: It shouldn't be very difficult but it could take some time depending on how much and in how many places the language files use and rely on some properties of the standard encodings. –  Ulrike Fischer Oct 6 '11 at 16:11