P→MPDFMDSystem documentation
Format referenceOpen converter ↗

Level 3 · Record and view

Last modified: (dates/navigation repair). Content current as of: (last retained content commit; content verification date unavailable). Original content last edited: . Current claims have not been freshly reverified.

PDFMD is the record; Markdown is a view

Recognition writes one canonical file containing text, structure, provenance, and explicit limits. A renderer then produces ordinary Markdown for a particular reading purpose.

On this page

Output package

One source record, one derived view, referenced assets

Figure 1 · Deployed package
document.pdfmdcanonical, UTF-8, extended Markdown
↓
human.mddefault derived reading view
+
assets/sha256-…referenced images or visual evidence

The canonical file stores each fact once. Text styling is encoded around that text; page and source relationships are attached as attributes or directives. Assets are content-addressed, so their names can be checked against their bytes.

Render-only route

Rendering does not repeat recognition

PDFMDcanonical input
→
Parsesyntax and extensions
→
ValidateIDs, references, assets
→
Apply profilechoose visible evidence
→
Markdowndisposable view

No PDF parsing, OCR, detector, model call, or human decision is required for a valid render-only operation. The same PDFMD can therefore produce several Markdown views without changing the recorded interpretation.

View settings

Physical page evidence can be shown or suppressed independently

SettingAvailable treatment
Page transitionsIgnore, stable placeholder, or an HTML comment plus horizontal rule.
Page headers and footersOmit, reproduce content, or insert a placeholder.
Printed page numbersOmit, reproduce content, or insert a placeholder.
DiagnosticsOmitted by default; a profile may insert stable diagnostic placeholders.

The settings are independent. For example, a reading view may omit headers while retaining page-transition markers for citation.

Conformance boundaries

Format validity is not source fidelity

A PDFMD file can be syntactically valid while containing a mistaken recognition decision. Conversely, a partial file can be valid because it explicitly records an unknown or failed region. The parser and validator check the record; evaluation against the source checks fidelity.

  • An unsupported required feature stops rendering.
  • Unsafe HTML, attributes, links, and package paths are rejected or sanitized according to the format rules.
  • Generated Markdown never replaces the canonical PDFMD file.
  • Unknown optional extensions survive a parse–serialize round trip.

Worked PDFMD example

Read one illustrative record line by line.

→

Compact PDFMD 1.0 reference

Syntax, directives, validation, rendering, security, and versioning.

→