From: Daniel Berger <djberg96@...>
Date: 2009-03-07T02:50:38+09:00
Subject: Re: Pure ruby iconv?



On Mar 6, 10:03 am, Charles Oliver Nutter <charles.nut...@sun.com>
wrote:
> James Gray wrote:
> > On Mar 6, 2009, at 9:59 AM, Daniel Berger wrote:
>
> >> Any suggestions? Or do I need to put a bounty out there for a pure
> >> ruby iconv now?
>
> > I sure hope you're planning to offer considerable more than $100 for
> > that effort.  This is no Zlib!
>
> That's for sure...Iconv is a brute. I'd love to see a pure Ruby version
> (and in this case we'd probably use it in JRuby as well), but it would
> mean a substantial amount of effort.
>
> Not that I'm trying to scare anyone interested in doing it, of course :)

Oh, sure, I could offer more money. But first I want to make sure
there isn't an alternative already out there.

BTW, it looks like Perl did it already with piconv, though I'm not
sure how much of it depends on features within Perl itself:

http://search.cpan.org/~jhi/perl-5.8.1/ext/Encode/bin/piconv

If there isn't a working alternative I'll probably handle this one
differently, i.e. I'd setup a separate rubyforge project for it in
advance, allow multiple developers (probably up to 3), hand pick the
developers if there are too many volunteers, set some milestones and
pay in increments.

However, it may not be necessary in the short term. I discovered,
thanks to my Solaris box, that the way Rails is handling the
redefinition of the transliterate method in inflector.rb is broken.
It's only checking the return value of Iconv.iconv, but not whether or
not an error is raised.

Time to file a bug report and patch.

Regards,

Dan