PDF Glossary

PDF/A is a PDF variant designed to guarantee a document remains readable in 10, 20, or 50 years. All fonts must be embedded, no external content allowed, the file is self-contained. Three families: PDF/A-1 (strictest), PDF/A-2 (with transparency and layers), PDF/A-3 (with file embedding).

In several European contexts (public archives, public tenders, electronic invoices), PDF/A is mandatory. For 95% of cases, PDF/A-2b is sufficient.

OCR transforms an image containing text (scan, photo) into usable text data — searchable, copyable, indexable.

Modern engines (Tesseract 5+, Google Vision, AWS Textract) use convolutional neural networks trained on tens of millions of pages, reaching 99.5%+ accuracy on clean French/English scans. On free handwriting, accuracy drops to 60-80% — still OCR's Achilles heel.

GDPR requires every company processing data of European citizens to respect principles of minimization, transparency, right to access, and erasure. It applies regardless of where the company is based.

GDPR compliance doesn't settle everything, though: as soon as data leaves the European Economic Area, specific rules apply (an adequacy decision such as the Data Privacy Framework for certified US companies, Standard Contractual Clauses, supplementary measures). And a provider subject to US law remains within the reach of the Cloud Act, wherever the data is stored.

The CLOUD Act (Clarifying Lawful Overseas Use of Data Act) lets US authorities, through legal process (a warrant or court order), require a provider subject to US law — including through its European subsidiaries — to disclose data in its possession or under its control, even if it's stored outside the United States. The affected user isn't necessarily informed.

Concretely: a PDF entrusted to a US company (for example Adobe, headquartered in California), or to a service that relies on a US provider for its hosting, may be concerned. To find out where a given service stands, read its privacy policy and subprocessor list. A tool that processes files in the browser, without uploading them, has nothing to disclose.

The eIDAS regulation (Electronic IDentification, Authentication and trust Services), in force since 2014, harmonizes electronic signatures in Europe. It defines three levels: simple (SES), advanced (AES), and qualified (QES).

QES is the only level strictly equivalent to handwritten signature before the law. It requires a certified device (smart card, national eID, certified app). SES (drawing your signature with a mouse) is legally valid but weakly probative — reserved for low-stakes uses.

DPI measures the pixel density of a digital or printed image. The higher the DPI, the more detailed the image — but the heavier the file.

Common values: - 300 DPI: the professional print standard - 150 DPI: enough for on-screen reading - 96 DPI: indistinguishable to the naked eye on screen (the limit of human perception)

PDF compression takes advantage of DPI: downsampling from 300 to 150 DPI typically divides the weight of images by 4, with no visible loss on screen.

When a PDF embeds a font, the whole font is included by default — including thousands of characters that are never used (Chinese ideograms, Cyrillic alphabets, mathematical glyphs). Subsetting keeps only the characters actually present in the document.

Typical saving: 30 to 200 KB per embedded font. On a document with 5 fonts, that's 1 MB saved with no loss of functionality. Almost every good PDF compressor subsets fonts automatically.

Factur-X (France) and ZUGFeRD (Germany) are two names for the same electronic invoice standard. A Factur-X invoice is technically a PDF/A-3 containing an embedded XML file in the CII (Cross Industry Invoice) format.

The PDF stays readable by people, the XML by accounting software. In France, electronic invoicing for B2B transactions is becoming mandatory on a phased rollout schedule, and Factur-X is one of the accepted formats.

pdf-lib is an MIT-licensed library for creating, reading, merging, splitting and modifying PDFs directly in JavaScript — no server needed. It's what lets PDFly's tools run "100% in the browser": your files never leave your device.

Typical performance: merging 10 PDFs of 5 MB takes 2-3 seconds on a recent MacBook. Because the whole document is loaded into memory, a very large file can overwhelm the tab, especially on mobile: that's why PDFly accepts files up to 100 MB and up to 50 files per operation.

The Schrems II ruling of the Court of Justice of the European Union (July 2020) invalidated the Privacy Shield, the agreement governing data transfers between the EU and the United States. The reason: US surveillance programs (FISA 702, EO 12333) allow access to European data without legal redress for Europeans.

After the ruling, transfers to the United States had to rely on other tools, such as Standard Contractual Clauses, backed where necessary by supplementary measures (encryption, pseudonymization…). Since July 2023, the Data Privacy Framework once again allows transfers to certified US companies; this new framework is still being challenged, however.

Redaction means permanently hiding sensitive information in a PDF — text, images, areas — in an irreversible way. Unlike a simple black rectangle drawn on top, true redaction removes the data from the source file.

Used for: GDPR compliance before sharing, publishing anonymized court rulings, protecting medical or legal data. Poorly done redaction (just a black rectangle) is a major risk: the data remains accessible by copying and pasting from the PDF.

AES-256 (Advanced Encryption Standard, 256-bit key) is the most widely used symmetric encryption algorithm for password-protecting PDFs. It's considered impossible to brute-force with today's computers.

A PDF protected with AES-256 can only be opened with the correct password. If you lose it, the content is permanently inaccessible (unless the password is weak enough for a dictionary attack). Prefer passwords of 12+ characters mixing upper case, digits and symbols.

A Tagged PDF contains a logical structure (headings, paragraphs, tables, image alt text) on top of its visual rendering. This structure lets screen readers (JAWS, NVDA, VoiceOver) navigate the document in logical order, skip decorative images and read tables correctly.

Required for accessibility compliance (WCAG 2.1 AA, RGAA, EN 301 549).

A linearized PDF ("Fast Web View") is restructured so the first page appears before the whole file has downloaded. This is essential for large PDFs (>10 MB) served on the web — a visitor won't wait 30 seconds for everything to download.

Linearization is invisible to the end user, but it changes the order of the objects in the file.

PDF metadata includes the author, creation date, originating software, keywords, and sometimes geolocation (on PDFs generated from a phone). It's stored in the file but not very visible — accessible via File → Properties in a PDF reader.

Before sharing a PDF, it can be useful to remove this metadata if it reveals sensitive information (a project's internal name, the version of proprietary software, the time zone of creation). Adobe Acrobat and some tools can "sanitize" the file.