From: Mike Durham <mdurham@...>
Date: 2006-10-23T11:25:10+09:00
Subject: Re: downcase part of a string

Wilson Bilkovich wrote:
> On 10/22/06, Mike Durham <mdurham@people.net.au> wrote:
>> James Edward Gray II wrote:
>> > On Oct 22, 2006, at 7:30 AM, ilhamik wrote:
>> >
>> >> Thanks Peter, it works fine.
>> >
>> > You missed Tim Bray's RubyConf talk.  According to him we should, never
>> > be using the case changing methods.  "Just don't do it!"  ;)
>> >
>> > James Edward Gray II
>> >
>> Why not? What reason did he give?
> 
> The problem is that proper upcasing and downcasing of characters is
> locale-dependent, not just encoding or language-dependent.
> 
> As examples, he mentioned that the uppercase version of accented
> characters varies from area to area in France.  Also, in Turkish,
> there are four different cases of 'i', not just two.. and which is
> correct depends on the jurisdiction.
> Determining the locale in a correct way is really, really hard. Tim
> Bray says it's basically impossible.  Also, all of these rules make
> any decent upcase/downcase function ruinously slow.
> 
> He shared a story about the original version of XML. At the time, it
> was case-insensitive.  The very first XML library was running horribly
> slow.  After profiling, they found that it was spending 90% of its
> time in the Java downcase routine.  After that, XML was made
> case-sensitive.
> 
Thanks Wilson, that explains everything. I'd never thought about 
problems like that.
Cheers, Mike