Encoding declaration in OPF and TOC?
 

I've made a lot of books with Sigil and sometimes I import them in Calibre. After having done that, I see that often accented characters are shown in the wrong way (e.g. Chinese characters instead of Latin ones) in the TOC and in the meta data.

About the first problem (TOC) I wrote to the programmer of Calibre and his response was "Bug fixed: When decoding NCX toc files, if no encoding is declared and detection has less that 100% confidence, assume UTF-8."

So I understand that the TOC should have an encoding declaration. Can this be added so that Sigil does that automatically? As I understand Sigil delivers perfect utf-8 but doesn't declare so.

Also about the second problem (errors in the meta data) I wrote to Calibre, and the answer was similar: stick an encoding declaration in the OPF.

Hence my similar question: Can Sigil add an encoding declaration to the OPF?

Thanks!

The XML standard states that an XML document without an "encoding" attribute in the XML declaration is encoded in either UTF-8 or UTF-16. If it is encoded in UTF-16, then it MUST have a Byte Order Mark. Therefore, if it doesn't have an encoding attribute and it doesn't have a BOM, it must be UTF-8.

In plain English, UTF-8 is the default character encoding for XML. I thought everyone knew that.

But I'll add the attribute, it can't hurt.

EDIT: And here's the source. Just invert the negatives. :)

I had the same opinion, but if a famous programmer says that I have to declare that my books are utf-8, I start doubting ;-)
Thanks! It will save a lot of trouble in many cases.

Oh if only everyone knew that and no one produced XML files encoded in encoding other than UTF-8 with no encoding declaration.

You should have said Kovid that this was causing you problems. It's no problem to add the encoding attribute. :)

But you really should fall back to the standard when byte stream fingerprinting isn't 100% sure of the encoding.

This is now in trunk.

You're marvellous, guys! :thanks:

The problem is that byte stream fingerprinting is almost never a hundred percent certain.