Archived copy. This page was recovered from the Internet Archive snapshot of /display/REQ/PDF to PDF-A Conversion Pre-Processor taken on 2011-10-08. The original wiki at wiki.opf-labs.org is being decommissioned.

PDF to PDF-A Conversion Pre-Processor

characterisation

Created by Carl Wilson on Sep 29, 2011 · last edited by Carl Wilson · on Sep 29, 2011 (view history) · 6 versions

Title	PDF to PDF/A Conversion Pre-Processing
Detailed description	It's unlikely that a batch PDF/A conversion utility could be written in just a couple of days: Because the PDF format is complex, the specification document is 700+ pages + extensions Conversion to PDF/A is a difficult problem, here's some of what's required All fonts embedded Well formed XMP metadata that's synched with the document information No external links No embedded objects (except other PDF/A files) Audio and video content is forbidden No encryption No LZW compression and more...... The approach chosen, characterise the test sample and the Isator PDF/A test corpus: Discover which sample files aren't PDF/A compliant, and why Compare results with characterisation of the test corpus Try to find the 20% of problems that cause 80% of the failures, and see if they can be fixed Create a tool that recognises the "hard" cases and highlights them as such to save effort At the British Library we already use Java utilities based on Apache PDF Box and JHOVE 1.6 for PDF characterisation. These tools can spot password protected and encrypted documents, embedded objects, and external links. They can also extract document information and XMP Metadata, and seemed to provide a reasonable fit. In reality the issues were very difficult: Many files fail for multiple reasons Some of the most common reasons
Solution Champion	Carl Wilson [email protected]
Corresponding Issue(s)	PDF to PDF-A conversion
Tool/code link	A link to code on Git hub or a corresponding myExperiment if applicable
Tool Registry Link	PDFBox
Evaluation	Notes from final presentations at the hackathon: CO: This problem is at the top of the list of challenges. A complete solution was described in DPC What's New as the CO's dream preservation tool. DEV: Work in progress characterisation toolset already identifies ~30% issues DEV: PDFTron identifies all problem cases! Looks to have been very thoroughly tested. DEV: PDFTron is a $600 tool