When request has Content-Type: text/xml;charset=UTF-8, the response is labelled as UTF-8 and any special characters are correctly encoded in UTF-8.
But when request has Content-Type: text/xml;charset=iso-8859-1, the response is labelled as ISO-8859-1 but the message body is still encoded as UTF-8.
See thread at http://marc.info/?l=axis-user&m=125062671421564&w=2 for more details.
Subject: Re: [AXIS2] Control character encoding used by axis2 server for
From: Andreas Veithen <andreas.veithen () gmail ! com>
Date: 2009-08-18 20:17:40
Message-ID: b67458760908181317v352bd106wd8fc9ae171728cc3 () mail ! gmail ! com
[Download message RAW]
The following log message is interesting:
org.apache.axiom.om.impl.MTOMXMLStreamWriter - Returning access to
the original output stream: java.io.BufferedOutputStream@14181418
It means that the xmlbeans databinding bypasses the normal StAX
XMLStreamWriter and writes part of the output directly to the
underlying byte stream. It is probably at that moment that things go
wrong. Can you please open a JIRA  and attach a sample AAR (or WAR)
that allows us to reproduce this problem?