[XeTeX] Polyglossia update

[Arthur Reutenauer]

On Sat, Oct 15, 2011 at 02:23:29PM -0500, Neal Delmonico wrote:
>                                                   One thing still
> bothers me about that whole affair.  I am working on several books
> involving Sanskrit and English and requiring hyphenation in both.
> None of the other books had that problem, as far as I know.  I
> wonder what was different about that book.

  It may be just chance: the wrong hyphenations such as n-ear, s-mall,
b-lissful that you mention in your mail from late September
(http://tug.org/pipermail/xetex/2011-September/021399.html) happened
because of the combination of three reasons: 1. they are allowed by the
hyphenation patterns for English; 2. they were allowed then because
\lefthyphenmin was (incorrectly) set to 1; and 3. TeX's line-breaking
algorithm just happened to find that breaking the line right there was
the best choice in the context of the paragraph (as hyphenation was
legal because of reasons 1 & 2).

  Changing one single word in a paragraph can completely change the
shape of the whole paragraph (remember, TeX always considers paragraphs
in their entirery before deciding where to break lines); hence it's very
hard to predict whether 3 will happen at all (short of actually
typesetting the text with TeX, of course).  It may thus be that you
didn't observe such incorrect hyphenations because 3 never occurred.
However, if you were typesetting long texts it's indeed odd that you
never saw anything like that, hence if you'd like to investigate the
matter more closely you can send other documents.

	Arthur (who knows the need to understand why things work when
		they do, just as much as why they don't when they don't :-)