Just closing the loop here a bit - another person here pointed out that it
seemed to depend on how you create the stream you feed to XmlTextReader in
the first place.

If you use File.OpenText() to get a StreamReader and then construct
XmlTextReader with a StreamReader, it appears to lock the encoding in place
and XmlTextReader will not respect the processing directive.

If you use File.OpenRead() to get a simple FileStream and use *that* to
construct XmlTextReader, the XmlTextReader is more responsive to what's in
the stream it's reading.

Thanks
-mark

[quoted text, click to view]

When the encoding of the XMLDecl and the encoding of the content
presented to the reader are different, then you will have problems.

[quoted text, click to view]

The documentation for File.OpenText( ) is clear about interpreting
the file as UTF-8,

http://msdn.microsoft.com/library/en-us/cpref/html/frlrfsystemiofileclassopentexttopic.asp

Even though the file MAY be encoded as iso-8859-1, doing this will
"present" the file's contents as UTF-8.

The encoding of the I/O StreamReader is paramount because remember,
XmlTextReader depends upon the StreamReader's Read( ) method(s).
The StreamReader is responsible for decoding from whatever bytes
are in the file to characters using it's encoding (it knows nothing about
XMLDecl).

[quoted text, click to view]

FileStreams can be binary, therefore choosing a FileStream gives the
XmlTextReader the option to read *bytes* instead of characters. It
then has something to say about what encoding it uses to perform this
translation.

[quoted text, click to view]

That's the XmlDeclaration's Encoding property. It won't appear as
an XmlProcessingInstruction. You could use this code to inject an
XMLDecl if one isn't already present,

if ( xml.FirstChild.NodeType != XmlNodeType.XmlDeclaration )
{
XmlDeclaration decl = xml.CreateXmlDeclaration( "1.0", "iso-8859-1", null);
xml.InsertBefore( decl, xml.FirstChild);
}

to set the XML Declaration if one doesn't exist. To read the
encoding off of an XmlDocument?'s XMLDecl,

string encodingStr = null;
if ( xml.FirstChild.NodeType == XmlNodeType.XmlDeclaration )
encodingStr = (XmlDeclaration)( xml.FirstChild).Encoding;
encodingStr = ( encodingStr == null ) ? "UTF-8" : encodingStr;

If the XmlDocument's FirstChild isn't of XmlNodeType.XmlDeclaration
then it doesn't have an XMLDecl. If there is no XMLDecl, or there is
one without an Encoding, then the encoding is UTF-8 by default.

In my experience, when the encoding on the XMLDecl matches the
encoding of the content, there are no problems.

I've tried producing a file to match your example like this,

- - - WriteOut.cs
using System;
using System.IO;
using System.Text;
using System.Xml;

public class WriteOutIso8859_1
{
public static void Main( )
{
FileStream fs = new FileStream( "iso8859_1.xml", FileMode.CreateNew);
StreamWriter writer = new StreamWriter( fs, Encoding.GetEncoding( "iso-8859-1"));
writer.WriteLine( "<?xml version='1.0' encoding='iso-8859-1'?>");
writer.WriteLine( "<root>");
writer.WriteLine( "\t<first>Hello World</first>");
writer.Write( "\t<second><![CDATA[");
writer.Write( new char[] { (char)0xED, (char)0xB3, (char)0xA8} );
writer.WriteLine( "]]></second>");
writer.WriteLine( "</root>");
writer.Flush( );
writer.Close( );
}
}
- - -

When I read this file in with the following code I have no problems.

FileStream fs = new FileStream( "iso8859_1.xml", FileMode.Open);
StreamReader sw = new StreamReader( fs, Encoding.GetEncoding( "iso-8859-1"));
XmlTextReader reader = new XmlTextReader( sw);
reader.MoveToContent( );
XmlNode node = xmlDoc.ReadNode( reader);


Derek Harmon