> > By the way, I have one proposal.
> > REXML::Encoding accords ICONV.rb priority over other encodings,
> > for example, EUC-JP.rb, ASCII.rb and so on.
> > 
> > If Iconv works well, this implementation is reasonable.
> > But, in some encodings, Iconv(libiconv) raises error and
> > REXML stops when parsing.
> This is strange, iconv works quite well. It only raises an error
> when
> the encoding given is wrong... And I think iconv is well tested...
> Are you sure the encoding of the file is really correct?
I should have explained the problem more precisely.

At least in Japanese encodings, libiconv uses useless encoding table.
(The word, 'useless', may be inappropriate. 
But, if we encode/decode Japanese string with libiconv, 
we sometimes meet error. )
This problem is caused by the differnce between 
CP932 and Unicode table of Microsoft and that of libiconv.

So, in Japan, patch for libiconv is provieded and widely used.
http://www2d.biglobe.ne.jp/~msyk/software/libiconv-patch.html
(Sorry, this page is written in Japanese.) 

As I said in the previous mail, 
this problem should be solved in libiconv. 
But, the author of the patch for libiconv said that
he couldn't merge his patch to libiconv. 

If we can't use the patch for libiconv, for example In ISP,
it is difficult to use REXML because of libiconv error.

In order to avoid this problem, I would like to add the option.


__________________________________
Do You Yahoo!?
Upgrade Your Life
http://bb.yahoo.co.jp/