Webinar September 2023 📍 Online

PDF Association TechConnect 2023 — Virtual: PDF/A-4 Deep Dive and PDF 2.0 Interoperability

The PDF Association TechConnect 2023 was a virtual developer-focused event dedicated to PDF/A-4 (ISO 19005-4:2020) technical details and PDF 2.0 interoperability testing results. Implementors shared real-world findings from library development and enterprise deployment of PDF 2.0 features.

Official event page →

Key takeaways

PDF/A-4 validation separates archival quality (PDF/A-4) from accessible tagging — unlike PDF/A-1a and PDF/A-2a, basic PDF/A-4 conformance does not require tagged content.
PDF 2.0 interoperability testing revealed page-level metadata and document part metadata as the most significant cross-vendor implementation gaps.
Associated file (/AF) handling on objects other than the document catalogue was inconsistent across implementations — a critical gap for ZUGFeRD/Factur-X workflows using PDF/A-4f.
CFF font subsetting edge cases, transparency compositing, and XMP version handling were identified as recurring implementation correctness issues.

The PDF Association TechConnect 2023 was a virtual developer-focused event held in September 2023, organised by the PDF Association. The TechConnect series is designed for PDF library developers, tool implementors, and technical architects rather than end-users or document professionals — the emphasis is on specification details, implementation experience, and interoperability testing results. The 2023 edition was dedicated to two related topics: PDF/A-4 technical implementation (what had changed from PDF/A-3, the new requirements, and what library developers needed to implement) and the results of ongoing PDF 2.0 interoperability testing among PDF Association member organisations. Both topics had direct implications for the developers of PDF generation, validation, and processing libraries.

PDF/A-4 technical deep dive

PDF/A-4 (ISO 19005-4:2020) was the primary technical topic at TechConnect 2023. Sessions provided specification-level detail that PDF Days conference sessions did not always reach: PDF/A-4 validation changes — The requirements for PDF/A-4 validation differ from PDF/A-1 and PDF/A-2 in several important ways. Implementors of validation engines (veraPDF, PAC, callas pdfToolbox) needed to update their PDF/A rule sets. Key changes: PDF/A-4 requires only that /MarkInfo /Marked true exists for tagged documents — it does not mandate tagged content for basic PDF/A-4 conformance (unlike PDF/A-1a and PDF/A-2a). The accessible accessibility requirement has been decoupled from the archival requirement. The colour management requirements in PDF/A-4 are aligned with the more flexible PDF/X-4 approach: an output intent is required only if the document contains device-dependent colour (same as PDF/A-2), but the acceptable output intent ICC profiles now align with the current ICC registry. Encryption is prohibited in all PDF/A-4 variants, as in all previous PDF/A standards. PDF/A-4f embedded files — The embedded file requirements in PDF/A-4f are technically similar to PDF/A-3 but with PDF 2.0 vocabulary: the /AFRelationship entry is now defined in ISO 32000-2 rather than as a PDF/A-3 extension. Implementation sessions reviewed the specific object structure differences between a PDF/A-3 embedded file and a PDF/A-4f embedded file. PDF/A-4e (engineering) — The engineering variant permits 3D artwork (U3D and PRC formats, as permitted in PDF 2.0) and rich media annotations. This variant was specifically designed to serve the engineering and manufacturing sector's need for a durable archive format that can carry 3D product data alongside 2D drawings. Sessions reviewed the 3D content requirements and the interaction with the archival constraints.

PDF 2.0 interoperability testing results

A major component of TechConnect 2023 was the presentation of PDF 2.0 interoperability test results collected through the PDF Association's interoperability testing programme. Participating organisations had run each other's PDF 2.0 output through their own parsers and validators, and the results revealed the real-world state of implementation consistency. What was tested — The test corpus covered: new object types introduced in PDF 2.0 (associated files, document part metadata, page-level metadata); new structure elements; updated encryption (AES-256 GCM); and correct handling of deprecated features that some tools still output. Consistent results — Core PDF 2.0 features that had been implemented for several years showed high consistency: cross-reference streams, object streams, basic encryption, and standard tagged PDF structure were handled correctly by all participating implementations. Problem areas — Several categories showed implementation divergence: Page-level metadata — PDF 2.0 introduced page-level metadata streams (each page can have its own /Metadata stream). Some parsers ignored page-level metadata; others reported it as an error; only a subset correctly used it. This was the most significant interoperability gap identified. Document part metadata — The /DPart (Document Part) hierarchy introduced in PDF 2.0 for variable data printing workflows had almost no cross-vendor implementation. It was present in the specification but not yet in most tools. Associated file handling — The /AF array for associated files (the PDF 2.0 normalisation of the PDF/A-3 mechanism) was handled inconsistently. Some parsers ignored /AF arrays on objects other than the document catalogue, while the specification permits /AF on pages, annotations, and other objects. Follow-up commitments — Participating organisations committed to addressing the identified divergences in their next releases. The PDF Association's interoperability testing programme was the direct mechanism driving these cross-vendor fixes.

Library implementation guidance

TechConnect events are distinguished by their direct technical guidance for library implementors — information that does not appear in the specification text but reflects implementation experience. CFF (Compact Font Format) font subsetting — Several sessions addressed edge cases in CFF font subsetting that had caused interoperability problems: specifically, the handling of CFF subroutines (shared glyph component data) in subset fonts, where incorrect subroutine remapping produced garbled glyphs in some rendering engines. Transparency group compositing — PDF transparency compositing (defined in the PDF specification's transparency model, originally derived from PostScript Level 3) has edge cases that are difficult to implement correctly. Sessions reviewed the specific blending mode and alpha compositing scenarios where implementations diverge, with reference to the test documents used in interoperability testing. XMP metadata version handling — PDF 2.0 requires XMP metadata to conform to the XMP 2022 specification (ISO 16684-1:2019). Several older PDF processors embedded XMP conforming to XMP 2004 or 2010 — parsers needed to handle both versions gracefully. ToUnicode CMap generation — Correct ToUnicode CMap generation (required for accurate text extraction from PDF) has many edge cases in non-Latin scripts: bidirectional text, complex script shaping (Devanagari, Arabic), ligature handling, and composite glyph decomposition. Sessions reviewed best practices for CMap generation in multilingual PDFs.

Frequently asked questions

What is the PDF Association TechConnect series?

TechConnect is a series of virtual technical events organised by the PDF Association, aimed specifically at PDF library developers, tool implementors, and standards-focused architects. Unlike the PDF Days conference series (which serves a broader audience including document professionals, managers, and standards users), TechConnect events go deep into specification details, implementation challenges, and interoperability testing results. Sessions often feature live demonstrations of specification edge cases, comparative results from multiple implementations, and direct discussion between developers of competing PDF libraries. TechConnect events are typically half-day virtual sessions; recordings are made available to PDF Association members.

What is the difference between PDF/A-4 and PDF/A-4f?

PDF/A-4 is the base conformance level for the fourth generation PDF/A standard (based on PDF 2.0). It requires: self-contained content (fonts embedded, no external dependencies), no encryption, no JavaScript, complete XMP metadata, and conforming colour management (output intent for device-dependent colour). PDF/A-4f extends PDF/A-4 with the permission to embed arbitrary file attachments, using the /AF (Associated Files) mechanism from PDF 2.0 with a required /AFRelationship declaration. PDF/A-4f is the successor to PDF/A-3 for workflows that need to embed machine-readable data (XML invoices, source files, datasets) alongside an archival PDF envelope. The "f" suffix stands for "files" — the embedded file capability is the sole addition over base PDF/A-4.

What are ToUnicode CMaps and why do they matter for text extraction?

A ToUnicode CMap (Character Map) is an entry in a PDF font dictionary that provides a mapping from the font's internal glyph encoding to Unicode code points. When a PDF viewer or text extraction tool reads text from a PDF content stream, it encounters glyph codes — integers that index into the font's glyph table. Without a ToUnicode CMap, those integers cannot be reliably converted to characters — the viewer must guess based on font encoding names (StandardEncoding, WinAnsiEncoding, etc.), which is often wrong for subsetted or non-standard fonts. With a ToUnicode CMap, every glyph code maps unambiguously to a Unicode character. Correct ToUnicode CMaps are required for: accessible text (screen readers need Unicode text); searchability; copy-paste from PDFs; text extraction for AI/LLM processing; and indexing. PDF generators that subset fonts must generate correct ToUnicode CMaps for the subset — only the glyphs actually used in the subset need entries.

Try it in your browser

The PDF tools and standards discussed at this event are available free — no account needed.

Edit PDF Compress PDF Open editor