[sword-devel] Greek dictionary - input needed
Chris Little
chrislit at crosswire.org
Wed Jan 21 00:56:34 MST 2009
Matthew Talbert wrote:
> What does ICU do that glib doesn't? See, for example,
> http://library.gnome.org/devel/glib/2.18/glib-Unicode-Manipulation.html.
glib appears to have a good selection of basic Unicode handling
functions. However, I don't see any reference made to UCA or CLDR. I'm
sure they're implementing the normalization algorithms correctly, but if
they aren't specifically using UCA and CLDR for collation, collation
won't be identical between glib and ICU. We really can't have some
segment of Sword implementers using ICU for collation and another
segment using glib if the two don't give the same results. (UCA is also
notoriously difficult to implement, so it's probably worthwhile to
contain ourselves to a single implementation if at all possible in order
to eliminate problems stemming from differing interpretations of the
algorithm.)
Another part of CLDR (which actually came from ICU's data) is a
multitude of date & time formats, calendars, and language names, all
translated into dozens of languages. Transliteration tables, another
element of CLDR, are likewise not mentioned in the cited glib doc.
ICU also has a significantly more advanced regex implementation than we
use currently in Sword, which would be a nice improvement. And its
implementation of the Bidi algorithm may help us solve some of the
issues we have regarding rendering of bidi languages like Hebrew,
Arabic, Farsi, & Syriac.
The ICU home page has a quick summary of its features and a list of many
products and organizations that use ICU: http://icu-project.org/.
--Chris
More information about the sword-devel
mailing list