[Podofo-users] PoDoFo cannot load some PDFs with XMP metadata

[Uli Z. <ul...@ri...>]

Hi,

I found that PoDoFo 0.9.2. cannot load some PDFs with XMP metadata.

In these cases, when PoDoFo::PdfMemDocument::Load() is called (I didn't try reading from a file), an exception is thrown with the error message "A number was expected but not found."

Upon closer inspection of the affected PDFs, I found that in all these cases, the XMP packet header

   <?xpacket begin='' id='W5M0MpCehiHzreSzNTczkc9d'?>

contains no character between the quotes of

   begin=''

This is allowed according to Adobe's XMP specification. In http://wwwimages.adobe.com/www.adobe.com/content/dam/Adobe/en/devnet/xmp/pdfs/cc-201306/XMPSpecificationPart1.pdf on page 11, it says:

> the character 􏰀 [the character between the quotes, U.Z.] represents the Unicode character U+FEFF used as a byte-order marker. The U+FEFF may be omitted from the begin="".


The current version of PoDoFo, however, stumbles as soon as there is no character between the quotes.

NOTE: Also contrary to Adobe's XMP specification (and the behavior of Adobe's XMP Toolkit SDK), PoDoFo seems to allow *any* character between the quotes of begin='', not only U+FEFF. However, I do feel that this is a *good thing* because it makes PoDoFo more fault tolerant. There seem to be several applications out there to write XMP data to PDFs, which don't get the packet header's U+FEFF character right. As a result, Adobe's XMP Toolkit SDK cannot read the XMP data of PDFs edited with these applications, while PoDoFo can. I consider this an advantage that should not be removed. (Specifically, there is a Java application from 2008 (still linked from creativecommons.org) that writes Creative Commons licenses in such an incorrect XMP format to PDF files. PoDoFo recognizes these licenses, Adobe's XMP Toolkit does not.)

Bye

Uli