Re: Intenational character sets

From: Frank van Bortel <fvanbortel_at_netscape.net>
Date: Mon, 31 Jan 2005 17:34:16 +0100
Message-ID: <ctlmjg$4ik$1@news3.zwoll1.ov.home.nl>

Ed Stevens wrote:
> On Fri, 28 Jan 2005 23:28:02 +0100, Sybrand Bakker
> <gooiditweg_at_sybrandb.verwijderdit.demon.nl> wrote:
>
>

>>On Fri, 28 Jan 2005 13:07:49 -0600, Ed Stevens <nospam_at_noway.nohow>
>>wrote:
>>
>>
>>>===============
>>>OMS does not have the database issue when the CHAR fields in the
>>>database use a BYTE semantic.  This is what was set when we developed
>>>OMS.  However, the CHAR fields were changed to use a CHAR semantic
>>>since french characters use more then one byte.  When this was changed
>>>to the CHAR semantic, that is when the issue arised. 
>>> 
>>>I tested changing the database fields to a BYTE semantic again and OMS
>>>does not have the issue.  From there I changed it back to use the CHAR
>>>semantic and recreated the mapping between the EJBs and database
>>>again.  Again the issue occurred.  There is a flag in Websphere that
>>>you can set for each field in an EJB to use the semantic CHAR or BYTE.
>>>I have made sure this flag is set to CHAR when I performed this test. 
>>> 
>>>I see this as a issue between the mapping of a String field in the EJB
>>>to a CHAR field in the database using the CHAR semantic.  
>>>=================
>>
>>
>>Translated
>>(and I don't know Websphere at all)
>>Websphere needs a 2-byte characterset (so anything UTF) to store
>>French characters (which is obviously quite ridiculous, as French
>>characters are in any Latin-1 compatible 1-byte characterset)
>>The database uses an one-byte characterset (or NLS_LANG on the
>>webserver is set incorrectly)

>
>
> Thanks for the reply. Immediately after my first post I started
> digging thru the docs. (I know that is usually backwards, but
> remember, my fundamental question was "where to start reading"?) After
> reading several relevant portions of "Database Globalization Support
> Guide" I've at least got some framework of understanding now. It
> looks like the database itself is well positioned for this, having bee
> created with characterset WE8ISO8859P1. My guess at this point is
> that their problem can be handled easily and elegantly with use of
> NLS_LANGUAGE on their end and they just need to be educated on that
> point. Unfortunately, at this point I still haven't gotten a response
> from the developers -- just their manages wanting to know when *I* am
> going to fix their problem!
>
> Film at eleven ..... ;-)
> Thanks.
>
> - Ed Stevens
>
> Cohn's Law: The more time you spend in reporting on what you are doing, the less time you have to do anything. Stability is achieved when you spend all your time doing nothing but reporting on the nothing you are doing.

Ed, you know you will never be able to store a 2 byte character in a 8-bits Oracle database (and West-European on top of it! :) ). So, whatever angle, you're going to loose on this one....

And... Sybrand is right: French characters do not take more than one byte - most of the time.

Call a managers meeting, tell 'em what the options are: 1) decide on a single byte character set, or 2) decide on a multi byte character set.

If 1), it's the nls_ settings and websphere..., if 2, you need to rebuild your db, as 8859P1 is not a subset of UTF.

-- 
Regards,
Frank van Bortel

Received on Mon Jan 31 2005 - 10:34:16 CST