Re: character converter (Re: current work)

[Aaron M. Renn]

Takashi Okamoto (address@hidden) wrote:
> Well, I want to add other japanese encodings like EUC-JP ,Shift_JIS...
> Does anyone implement them? If not, I would try that.

Thanks for your contribution.  Implementing these character encodings 
would be very helpful.  There seemed to be some difference in how
character encodings were handled between Classpath and the Cygnus
code.  We should probably ensure that there is a common mechanism
going forward so that we don't end up with rework.  I'm particularly
interested to see how the framework will handle complex character sets.
 
> I have following two opinions for character converters that are iconv
> converter and Java converter.
 
Ah, I see you've already thought of this.  The easiest way for now is
probably to use iconv, whatever that is.  If necessary, we can
distribute the files from it ourselves if we need to build outside of
a glibc environment.  However, somebody ought to look at a quick
performance benchmark.  The gcj code is designed to call native C++
using CNI, which is a high performance interface.  Our previous
experience with Japhar indicates that standard JNI interface can have
a substantial performance penalty versus native Java code.  For a
core function such as character conversions, we don't want to end up
with a solution that has poor performance.

-- 
Aaron M. Renn (address@hidden) http://www.urbanophile.com/arenn/