You may want to cross post this with the facelets list as well. I am
doing so in this reply...
On 10/31/06, Randahl Fink Isaksen <[EMAIL PROTECTED]> wrote:
I finally found out why the combination of Facelets 1.1.11 and MyFaces
1.1.4 makes it impossible to serve UTF-8. The problem is that they both
expect each other to provide the encoding. First the FaceletViewHandler
tries to make MyFaces decide which encoding to use by doing this:
String encoding = null;
ResponseWriter writer = renderKit.createResponseWriter(
NullWriter.Instance, contentType, encoding);
encoding = getResponseEncoding(context,
writer.getCharacterEncoding());
For now, just notice here that createResponseWriter() is invoked with an
encoding of null. In MyFaces the default render kit is HtmlRenderKitImpl
and when Facelets call createResponseWriter() with a null encoding
MyFaces responds by doing this:
if(characterEncoding==null)
{
characterEncoding = HtmlRendererUtils.DEFAULT_CHAR_ENCODING;
}
So MyFaces essentially says that if Facelets does not state explicetly
which encoding it wants it just uses its default encoding, and
DEFAULT_CHAR_ENCODING is in MyFaces equal to ISO-8859-1.
Now, back to the first code snippet - in the last line of the shown
FaceletViewHandler code above a method called getResponseEncoding() is
invoked and it contains this code:
if (encoding == null) {
encoding = "UTF-8";
This code will not have any effect because Facelets already invoked
MyFaces with a null encoding which made MyFaces fall back to ISO-8859-1
and thus the encoding is not null anymore.
So is this a MyFaces bug or a Facelets bug? In the code above Facelets
deliberately chooses not to decide for a specific character encoding
thereby expecting that
1. the JSF implementation provides the developer with means for
specifying which character encoding he wants and that
2. the render kit provided by the JSF implementation will return this
desired character encoding
This is in my opinion a wrong assumption - nothing in the JSF
specification 1.1 says this is so. On the contrary the specification
explicitly states that the encoding parameter is required when you
invoke createReponseWriter, quote: "... the required value for the
characterEncoding parameter for this method..." (section 8.1). So
invoking createResponseWriter() with a null encoding is a violation of
the specification, and thus I think this is a serious Facelets bug.
I sincerely hope the Facelets developers will comment on this. Thank you.
Randahl
