"use strict"; const {Fragment: _Fragment, jsx: _jsx, jsxs: _jsxs} = arguments[0]; const {useMDXComponents: _provideComponents} = arguments[0]; function _createMdxContent(props) { const _components = { a: "a", code: "code", h2: "h2", li: "li", p: "p", pre: "pre", span: "span", ul: "ul", ..._provideComponents(), ...props.components }; return _jsxs(_Fragment, { children: [_jsx(_components.p, { children: "If you had asked me what a PDF was before this job, I would have said \"a file for stuff that does not change.\" That was the whole of my technical opinion. They print. They open in every browser. You attach them to emails." }), "\n", _jsx(_components.p, { children: "A year into working on PDF tooling at Apryse, I have come around on that opinion in a big way. The PDF format is one of the more quietly ambitious pieces of software engineering I have seen up close. It is a container for text, vector graphics, bitmaps, fonts, embedded scripts, forms, annotations, digital signatures, accessibility metadata, and honestly anything else a working group ever wanted to cram into a document. It survived web 1.0, web 2.0, mobile, cloud, and every flavor of \"the future of documents\" pitched over the past thirty years. Somehow it is still the most interoperable document format on the internet." }), "\n", _jsx(_components.p, { children: "I want to walk through how we got here, what is actually inside a PDF, and why \"just use PDF\" is both the best and worst answer to a lot of modern document questions." }), "\n", _jsx(_components.h2, { children: "From PostScript To PDF 1.0" }), "\n", _jsx(_components.p, { children: "You cannot understand PDF without a stop at PostScript. PostScript was Adobe's page description language in the 80s. It was a Turing complete programming language that a printer would execute to produce a page. You sent a PostScript file to the printer, the printer ran it, paper came out. Revolutionary at the time. Also a little unhinged, since a buggy PostScript program could infinite loop your printer." }), "\n", _jsx(_components.p, { children: "PDF shipped in 1993 as Adobe's answer to, \"what if we took the useful parts of PostScript and stripped out the programming language.\" PDF borrowed the drawing model, paths and fills and strokes and text positioning, and dropped the imperative stateful parts. A PDF describes the same visual output a PostScript program would produce, but it does it declaratively." }), "\n", _jsx(_components.p, { children: "That choice, declarative over imperative, is the reason PDF won. You cannot infinite loop a PDF renderer. You cannot hang a viewer the way early printers could hang on bad PostScript. Deterministic rendering with a bounded execution model is why every browser, every OS, every mobile platform ships a PDF viewer today." }), "\n", _jsx(_components.h2, { children: "What Is Actually Inside A PDF" }), "\n", _jsx(_components.p, { children: "People who have not worked with the format up close tend to be surprised here. A PDF is not a tightly packed binary blob. It is closer to a small filesystem wrapped in an envelope. Open one in a hex editor and you find:" }), "\n", _jsxs(_components.ul, { children: ["\n", _jsx(_components.li, { children: "A header identifying the PDF version" }), "\n", _jsx(_components.li, { children: "A body full of objects (text chunks, images, fonts, scripts) each with an object number" }), "\n", _jsx(_components.li, { children: "A cross reference table telling the parser where every object lives in the file" }), "\n", _jsx(_components.li, { children: "A trailer pointing at the cross reference table and the document root" }), "\n"] }), "\n", _jsx(_components.p, { children: "The objects reference each other. A page object points to its content stream and its resources. A content stream points to fonts and image objects. Fonts can be embedded subsets of TrueType or OpenType files. It is a small object graph on disk. Parsing a PDF is traversing that graph." }), "\n", _jsx(_components.pre, { children: _jsxs(_components.code, { className: "code-highlight", children: [_jsx(_components.span, { className: "code-line line-number", line: "1", children: "%PDF-1.7\n" }), _jsx(_components.span, { className: "code-line line-number", line: "2", children: "1 0 obj\n" }), _jsx(_components.span, { className: "code-line line-number", line: "3", children: "<< /Type /Catalog /Pages 2 0 R >>\n" }), _jsx(_components.span, { className: "code-line line-number", line: "4", children: "endobj\n" }), _jsx(_components.span, { className: "code-line line-number", line: "5", children: "\n" }), _jsx(_components.span, { className: "code-line line-number", line: "6", children: "2 0 obj\n" }), _jsx(_components.span, { className: "code-line line-number", line: "7", children: "<< /Type /Pages /Kids [3 0 R] /Count 1 >>\n" }), _jsx(_components.span, { className: "code-line line-number", line: "8", children: "endobj\n" }), _jsx(_components.span, { className: "code-line line-number", line: "9", children: "\n" }), _jsx(_components.span, { className: "code-line line-number", line: "10", children: "3 0 obj\n" }), _jsx(_components.span, { className: "code-line line-number", line: "11", children: "<< /Type /Page /Parent 2 0 R /MediaBox [0 0 612 792]\n" }), _jsx(_components.span, { className: "code-line line-number", line: "12", children: " /Contents 4 0 R /Resources << /Font << /F1 5 0 R >> >>\n" }), _jsx(_components.span, { className: "code-line line-number", line: "13", children: ">>\n" }), _jsx(_components.span, { className: "code-line line-number", line: "14", children: "endobj\n" })] }) }), "\n", _jsx(_components.p, { children: "That is a minimal PDF. The R references mean \"object 2, generation 0\" and so on. Once you see it, the format is surprisingly readable, which is a delight if you have ever tried to spelunk through an old .doc file." }), "\n", _jsx(_components.h2, { children: "PDF 2.0, The Spec Nobody Talks About" }), "\n", _jsx(_components.p, { children: "Most of the PDF world still thinks of PDF 1.7 as the standard. There is a reason. It is what Adobe Reader treats as its baseline. It is what ISO 32000-1 codified in 2008. It is what 99% of PDFs in the wild target. PDF 2.0 (ISO 32000-2) came in 2017 and got an update in 2020, but adoption has been slow." }), "\n", _jsx(_components.p, { children: "The 2.0 spec cleans up a lot of legacy baggage. Older encryption schemes get deprecated. Rendering ambiguities get tightened. Better support for Unicode text extraction and structured accessibility content." }), "\n", _jsx(_components.p, { children: "Here is the thing I did not appreciate before this job. Changing a standard with three decades of existing content depending on it is brutal. You cannot break backward compatibility. You cannot force millions of old PDFs to upgrade. So 2.0 adoption is a gentle curve on purpose, and tools like ours have to treat 1.x and 2.0 as equally first class." }), "\n", _jsx(_components.h2, { children: "Fonts Are Their Own Saga" }), "\n", _jsx(_components.p, { children: "If you want to know why PDF parsing is hard, look at fonts. A PDF can embed:" }), "\n", _jsxs(_components.ul, { children: ["\n", _jsx(_components.li, { children: "Type 1 fonts (Adobe's original vector format, mostly legacy now)" }), "\n", _jsx(_components.li, { children: "TrueType (the one most people know from the 90s)" }), "\n", _jsx(_components.li, { children: "OpenType (the extended format Microsoft and Adobe both settled on)" }), "\n", _jsx(_components.li, { children: "CFF fonts (Compact Font Format, used inside OpenType and Type 1-in-PDF)" }), "\n", _jsx(_components.li, { children: "Composite fonts with CJK specific encodings" }), "\n", _jsx(_components.li, { children: "Subsetted fonts containing only the glyphs the document uses" }), "\n"] }), "\n", _jsx(_components.p, { children: "Each of these has quirks. Each can be embedded with different character encodings. Some PDFs include ToUnicode CMaps that map glyphs back to Unicode code points for text extraction. Some skip that, which is why copy paste from certain PDFs gives you nonsense." }), "\n", _jsx(_components.p, { children: "A surprising amount of my past year has been chasing font edge cases. You would think fonts were a solved problem. They are not. Somewhere in the world is a font that breaks any PDF library you care to name." }), "\n", _jsx(_components.h2, { children: "Forms, Signatures, And The Rise Of Interactive PDF" }), "\n", _jsx(_components.p, { children: "PDF started as a purely visual format. Around PDF 1.2 it began picking up interactivity. Form fields you could fill out. JavaScript that ran on open. Annotations. Digital signatures. Modern PDFs are as much applications as they are documents." }), "\n", _jsx(_components.p, { children: "Good news, these features make PDF much more useful for business workflows. You sign a contract in a PDF and the signature can be cryptographically verified without any external system. You have forms that validate input on the client. You embed fillable templates that render the same on every device." }), "\n", _jsx(_components.p, { children: "Bad news, a PDF with JavaScript is a PDF that can execute code. That has been an attack surface forever. Modern viewers disable PDF JavaScript by default, including the built in ones in Chrome and Edge. Customers ask all the time why their PDF JavaScript \"does not work.\" The answer is usually that the browser is protecting the user from code it cannot trust, because it does not know if the script is helpful or hostile." }), "\n", _jsx(_components.h2, { children: "Accessibility And PDF/UA" }), "\n", _jsx(_components.p, { children: "One thing I have grown to respect is how seriously the modern PDF spec takes accessibility. Tagged PDF, the underpinning of PDF/UA, adds a structure tree on top of the visual content. The tree describes what each chunk of the document is, heading, paragraph, list item, table cell. A screen reader uses this structure to read the document meaningfully instead of spilling positioned text fragments." }), "\n", _jsx(_components.p, { children: "I have a separate post in me about this, so I will not go deep. Short version, a well tagged PDF is as accessible as a well structured HTML page. An untagged PDF is a mess for assistive tech. The tooling to make PDFs accessible is getting better every year, and it is one of the places where working on PDF tooling has felt meaningful." }), "\n", _jsx(_components.h2, { children: "The Browser Angle" }), "\n", _jsxs(_components.p, { children: ["Hard to talk about modern PDF without the browser. Every major browser now ships a PDF viewer. Chrome uses a fork of ", _jsx(_components.a, { href: "https://mozilla.github.io/pdf.js/", children: "pdf.js" }), ", Firefox uses pdf.js proper, Safari rolls its own. These handle 90% of PDFs 90% well. They struggle with unusual fonts, certain form types, digital signatures, and anything JavaScript heavy."] }), "\n", _jsxs(_components.p, { children: ["The interesting effect is that these built in viewers raised the baseline expectation for a generation of users. People now expect to open a PDF in their browser. They expect to annotate, highlight, fill a form in place. Tools like our ", _jsx(_components.a, { href: "https://docs.apryse.com/documentation/web/", children: "WebViewer" }), " exist because the built in viewers are the floor, not the ceiling. Real business use cases need a lot more than the floor."] }), "\n", _jsx(_components.h2, { children: "Why PDF Is Not Going Away" }), "\n", _jsx(_components.p, { children: "People keep predicting the end of PDF. \"We will all use HTML forms instead.\" \"We will use structured data instead of documents.\" Every year another proposal shows up claiming to obsolete PDF. Every year, PDF remains the format people reach for when they need a document to survive being emailed, printed, reopened in five years, and still look the same." }), "\n", _jsx(_components.p, { children: "That is the thing about PDF. It solves a boring, human scale problem. \"I want this to look the same everywhere, forever.\" Nothing else does it as well. Word documents reflow. Web pages break. Images lose fidelity. PDF sits there looking the same on every device for decades." }), "\n", _jsx(_components.p, { children: "A year in, I think PDF is underrated. Boring on the surface, fascinating once you crack it open, one of the more successful file formats in computing history. Every major OS ships a PDF viewer, and it is not because PDF is trendy. It is because PDF works." }), "\n", _jsx(_components.p, { children: "I still have plenty to learn about the format. Years of spec left in there, corners I have not touched. But if you had told me a year ago that I would find PDF tech interesting, I would have laughed at you. I was wrong. It is a rabbit hole, and a good one." })] }); } function MDXContent(props = {}) { const {wrapper: MDXLayout} = { ..._provideComponents(), ...props.components }; return MDXLayout ? _jsx(MDXLayout, { ...props, children: _jsx(_createMdxContent, { ...props }) }) : _createMdxContent(props); } return { default: MDXContent };

The Evolution of PDF Technology: Reflections After a Year at Apryse