Turkish Character Problem Using Database Link

Question

Hi all,
I have a link created to a 10g XE database, from a 9i database. The link works fine for everything, but there is a problem with Turkish characters. Consider following query;
INSERT INTO xxer_char_deneme_brk@teta -- obviously teta is the link from 9i to 10g XE
SELECT urun_adi, --Atkı - CQS - 00 Beden
'ĞÜŞİÖÇI','ğüşiöçı',
URUN_UZUN_ACIKLAMA --%35 VİSKOZ %29 YÜN %20 POLYAMID %8 ANGORA %8 KAŞMİR
FROM xxer.xxer_stok --9i table
WHERE urun_adi LIKE '%Atk%';
In this query as you can see I insert two column values manually, and two from a table in my production database(9i). Also I show the correct values of columns as comments. When I run this query those columns are inserted to the remote database(10g XE) as 'Atki - CQS - 00 Beden' and '%35 VISKOZ %29 YÜN %20 POLYAMID %8 ANGORA %8 KASMIR', but the manual values are inserted without any conversion. As you can see only the values selected from the table change. For example İ is converted to I, Ş to S, Ğ to G and so on.
Now first I thought of NLS parameters, but both databases have Turkish language and territory settings. 9i runs on AIX 5.3, 10g runs on Windows server 2008.
Any help will be appreciated.
Thanks
Burak

Answer 1

Pl see these MOS Docs
Changing the NLS_CHARACTERSET to AL32UTF8 / UTF8 (Unicode)          [Document 260192.1]
AL32UTF8 / UTF8 (Unicode) Database Character Set Implications          [Document 788156.1]
The easiest option would be to export from the current XE database, reinstall the AL32UTF8 version of XE, and perform an import.
http://download.oracle.com/docs/cd/E11882_01/server.112/e10729/ch11charsetmig.htm#CEGCHGIJ
HTH
Srini