It turns out that there are a lot of encoding issues in the application.
Valfréjus is the example.
the é was encoded as C3 83 C2 A9. This turned out to be twice encoded
utf8. so in perl, I need to decode it twice. This seems to work but not
in all cases. Unfortunately. It is a very confusing situation anyway.
Since this application works in a client / server environment, we need
to see where the problems start.
At the server, we have a database, an XML interface, a webserver. I
don't have access to that part, because it third party software.
At the client, we have Perl, SOAP::Lite.
Now I am wondering, does SOAP::Lite any character conversion? Or just
pass it through without touching the data?
I believe it works that way, because in the HTTP layer, I can see the
wrongly coded parts.