Character encoding in resource bundle files

Question

I am developing a multilingual site, using resource bundle files. The problem I'm having is that, while the jsp file is save in UTF-8, and the contentType="text/html; charset=UTF-8", the foreign characters in the bundle files don't display properly unless I save them as "windows-1252". If I copy the text and paste it into the jsp file, it displays correctly.
How can I make the text from the bundle files display the same as the text in the jsp document itself when both files are saved as UTF-8? I'm surprised it doesn't do this by default.

Answer 1

Because resource bundles (the PropertyResoureceBundle class in particular) doesn't know anything about JSP or HTML page encodings. All that class knows is that it needs to load a properties file from the local (server-side) file system. (BTW, if you are using Struts MessageResources, it works the same way). Anyway, it doesn't read the file every time you call for a resource property String, it reads it 1 time, and 1 time only when you call new ResourceBundle().
The API docs (in java.util.Properties) specified that properties files are read using the ISO8895-1 encoding, actually, not the Windows-1252 or whatever the OS defines. So if you store text that's using a different encoding then ISO8895-1, it's going to read the chars incorrectly.
native2ascii converts things to ASCII using Unicode escapes, as I said. (Although, maybe it saves as ISO8895-1, I forget..., or maybe you should specify the output encoding.... you'd have to read the docs on that tool, cuz there's no command-line help.) This is sort-of an exception to the ISO8895-1 thing, in a sense, as all ASCII chars are the same as the first 127 chars of ISO8895-1, so it doesn't matter really.
So in short, it doesn't assume it's in the default Windows encoding, it assumes it's in ISO8895-1. But since the API docs specifically say so, you should abide by those rules.
As for "getting around the problem", yes, it does, actually. See, whatever it reads in that Unicode escaped ASCII text is converted properly to Strings in Java, which are 16-bit Unicode characters, which cover nearly every character set around (English, Chinese, Japanese, Arabic, Cyrillic, etc). When it's in that form, the chars are the Unicode chars and that's it. That's all they need to be. Encodings are only important when writing (and reading back) to files or other byte streams. You have to read in with the same encoding it was written under, otherwise the conversion is not going to be correct (with a few exceptions, but not many).
So that creates a Unicode String in Java, and when you write that to a JSP page or servlet, the output writer used knows from the content type you specified in the page declaration (UTF-8 is typically recommended because it's an efficent way to deal with Unicode chars, particularly in HTML, where most of the HTML tag content is still ASCII text), so it encodes the strings in that encoding.
Get it?