From: Trent Shipley <tshipley@deru.com>
Reply-To: tshipley@deru.com
To: pgsql-sql@postgresql.org
Subject: Re: a "catch all" type ... such a thing?
Date: Sat, 10 Sep 2005 20:54:00 -0700
User-Agent: KMail/1.8
References: <20050908181850.G1104@ganymede.hub.org>
	<200509102043.31876.peter_e@gmx.net>
In-Reply-To: <200509102043.31876.peter_e@gmx.net>
MIME-Version: 1.0
Content-Type: text/plain;
  charset="iso-8859-1"
Content-Transfer-Encoding: 7bit
Content-Disposition: inline
Message-Id: <200509102054.00504.tshipley@deru.com>

On Saturday 2005-09-10 11:43, Peter Eisentraut wrote:
> Marc G. Fournier wrote:
> > Are there any data types that can hold pretty much any type of
> > character? UTF-16 isn't supported (or its missing from teh docs), and
> > UTF-8 doesn't appear to have a big enough range ...
>
> UTF-8 has exactly the same "range" as UTF-16.  In any case, the UTF-8
> encoding in PostgreSQL is probably your best choice, unless you want to
> dig into the weirdness that is MULE_INTERNAL.

The 8.1 beta documentation says that UTF-8 in earlier versions of Pg only 
covered the first 16 bits of Unicode.  Unfortunately "pure" Unicode uses 32 
bits and (according to my Unicode Demystified) needed at least 21 (?) bits to 
represent all the code points available in Unicode 3.x.  (I think Unicode is 
now in 4.x.)  

This means that the code space supported by Pg 8.0 is technically too small.  
It shouldn't matter though, unless you are working with Chinese or a private 
character set.