Re: ICU

[Bruno Haible <bruno clisp org>]

Behdad Esfahbod wrote:
> I still have not looked at ICU (would do soon), but I strongly
> believe that we can borrow a lot of code from it.

Not sure about that, because
  1. Half of ICU is in C++, and some of the C APIs use C++ implementations.
  2. ICU is the spearhead in i18n research. Mark Davis and others actively
     hack it, to find out the best algorithms here and there. A project
     which just implements the algorithms from the specification will
     be slimmer.

> Or are you
> still suggesting we do not start this project and use ICU
> directly?  Can you persuade Bruno, for example, to depend on ICU
> in gettext?

You cannot convince me, because I want the libraries that I rely on to
be understandable. Whereas alone the code of ICU's collation code is
300 KB of source. Even glibc and the Linux kernel are easier to understand
than that.

Bruno