Re: Bug in Mozilla parser

Vinu wrote:

> Hi,
> 
> I am using the mozilla parser originally written by JamesClark.
> I set the encoding to latin_1.
> 
> My xml buffer contains some characters greater than Ascii code 128. But
> these are in the CDATA sections.

Doesn't matter. A character in the document is a character in the
document, even if it's in a CDATA section or a comment.

> Now my XML buffer is clearly in latin_1 or ASCII encoding. 

No, specifying <?xml version="1.0" encoding="iso-8859-1"?> just says
that you *claim* it's all Latin-1; it doesn't guarantee or check it.

> So when the 
> parser encounters these characters(ASCII code > 128), it uses two bytes
> to represent it rather than the one byte used for latin1.

You mean the parser encounters two bytes where you were expecting your
document to use only one?

> This is obvioulsy a UTF-8 representation.

Possibly.

> But i have clearly set the encoding as latin1. Then why is this
> happening.

Whatever process generates your XML (hand-edit? program? download?)
is storing the character in a character encoding other than ISO 8859-1.

You need to find out where the data is coming from and fix it.

> Also isnt the parser supposed to ignore anything in the CDATA section.

No, a CDATA section only stops the parser scanning for markup characters.
The content still has to be in the document character set.

> Is this a well known bug. 

No, but it is a common problem, usually due to the XML document being
stored in a different encoding to the one you are expecting.

> Could somebody please suggest something on 
> this.

If you're using a Unix-based system (Mac, Linux, etc), and if you are
positive that the offending character has a representation in ISO 8859-1,
use the iconv program to force the document to be stored as such, eg

$ iconv -t iso-8859-1 myfile.xml >newfile.xml

If the character has no representation in ISO 8859-1 you'll get an
error message, and you'll have to use UTF-8 or something else instead.

I don't know how you do this under Windows, though.

///Peter