Hi Thomas,

I would have loved to give some code, but there is nothing special in
the code. We are using a set of input methods that enables us to
provide input in Indian languages. These are in the form of jar files,
which specify the input characteristics of various Indian Languages.
All these Jar files lie in extention directory inside the JRE.

Now if i don't explicitly set the file encoding, there is no problem
locally. But when the same data comes on the Java socket (Application
is distributed, needs to send messages in local languages some times)
then the data is all garbled, questions marks or usual rectangles, in
the GUI.

In the system menu, a item gets installed automatically when an app is
run giving access to various input methods.

In our application when an item is selected in the menu, we just change
the input method, and which is taken care of those classes lying in the
extention.

I have a hunch that some other setting is required which will dictate
the display of widgets.

I have two installations
1.  1.4.2_08 (Jars in extention)
2. 1.5.0_03 (No Jars)

With [1] the description remains same, but with [2] the same app is ok

Thanks everyone for the replies, but unfortunately I am still in the
haze.

Normally you specify the encoding in your i/o code. See
http://mindprod.com/applets/fileio.html
for how to do it.

Yes, the constructors
  InputStreamReader(InputStream)
  OutputStreamReader(OutputStream)
  FileReader(...)
  FileWriter(...)
  -and may be some more?-
do so.

But furthermore the constructors
  String(byte[])
  String(byte[], int, int)
and the methods of String class
  byte[] getBytes()
  void getBytes(int, byte[], int, int)
depend on the default encoding, too.

I recommend to avoid these constructors/methods altogether.
Most of them have sister constructors/methods which take an encoding
name as an additional String argument.