History Lesson: What Is a .png File? The Format Built to Replace GIF

What is a .png file?

A .png file is an image stored in the Portable Network Graphics format. Spell the idea plainly: it is a lossless compressed picture (the decompressor rebuilds the exact pixels the encoder stored), built to travel the early web without the patent baggage that hit GIF. The official pronunciation is “ping.” A recursive joke — “PNG’s Not GIF” — stuck in developer lore beside the sober name.

On disk a PNG is a short signature followed by chunks: named blocks that carry the header, palette, compressed pixels, text comments, and an end marker. That design is why PNG could grow optional features without pretending every file was a multi-frame cartoon. The everyday search “what is a .png file” is usually someone staring at a sharp screenshot, a logo with a soft shadow, or a diagram that looks “too big” next to a photo JPG — and wondering whether something is wrong. Usually nothing is wrong; the format is doing a different job. For the GIF story this format answered, see History Lesson: What Is a .gif File?.

Why is my PNG so much bigger than a JPG?

Lossless versus lossy is the whole shock. PNG keeps pixel values recoverable after decompression. JPEG (the everyday JPG photo path) throws away high-frequency detail the eye rarely notices in continuous-tone photos, which is why a vacation snapshot is often tiny as JPG and stubbornly large as PNG. A Windows screenshot, a UI mock, or a scanned line diagram has sharp edges and flat color — exactly where lossless wins and lossy ringing looks ugly.

Transparency adds bytes too. GIF offered one transparent palette index — a hard on/off “hole.” PNG can store an alpha channel: a separate transparency sample per pixel, so soft shadows and glass edges composite cleanly over any background. That power is why designers loved PNG and why a translucent export is larger than a flat JPG of the same width and height.

The Unisys LZW fight and why PNG started

GIF had become the early web’s workhorse for logos and simple graphics. Its compression used LZW, and Unisys held a key patent. After talks with CompuServe, a late-December 1994 announcement made clear that GIF-supporting software could mean royalties. Usenet erupted in early January 1995. This lesson does not invent fee schedules; the cultural fact is enough: authors who thought GIF was free to implement suddenly faced a licensing problem. Protests such as “Burn All GIFs” and the push for a clean replacement followed. The prior GIF history lesson covers CompuServe, GIF87a / GIF89a, and the patent timeline in more depth — PNG is the sequel.

Thomas Boutell posted the first draft of what became PNG — then called PBF (Portable Bitmap Format) — to graphics and compression newsgroups on 4 January 1995. Within about a week the volunteer circle had proposed most of the lasting ideas: delta filtering before compression (Scott Elliott, Mark Adler, and others), deflate compression (Tom Lane, the Info-ZIP community, and many more), truecolor depth, internal CRCs (Greg Roelofs), and the PNG name itself (Oliver Fromme). Boutell steered drafts as a benevolent dictator when consensus stalled. By Draft 9 (7 March 1995) the format was effectively frozen; later drafts mostly clarified.

How the bytes work: signature, chunks, DEFLATE, filtering, Adam7

Every PNG begins with an eight-byte signature so readers can recognize the file even when the extension is wrong. After that comes a stream of chunks. Spell chunk as a self-describing block: a length, a four-character type name, a data payload, and a CRC (cyclic redundancy check) — a checksum that detects corruption in the type and data. Critical chunks every decoder must understand include IHDR (image header: width, height, bit depth, color type, and flags), one or more IDAT chunks (image data: the compressed pixels), and IEND (image end). Optional chunks can carry a palette, simple or full transparency, gamma hints, text, or other metadata without breaking older readers that skip unknown ancillary types.

Pixels are not stuffed raw into IDAT. First comes filtering: each scan line can be rewritten as differences relative to neighboring bytes so repeated patterns compress better — think of it as a reversible pre-pass, not a second image format. Then the filtered bytes are compressed with DEFLATE (the LZ77-plus-Huffman method also used in gzip / ZIP) through the zlib format. Jean-loup Gailly and Mark Adler shaped zlib for this era; Guy Eric Schalnat led early libpng work so applications did not each reinvent the decoder. Specs and libraries matured together through 1995.

Interlacing is optional. Method 0 stores rows top-to-bottom as usual. Method 1 is Adam7, named for Adam M. Costello: seven passes over an 8×8 pattern so a coarse preview appears before the full download finishes — the progressive-display habit dial-up browsers needed. Decoders must read interlaced files even if they only show the final image.

W3C Recommendation, RFC 2083, and ISO/IEC 15948

Standardization followed implementation. The World Wide Web Consortium (W3C) issued PNG 1.0 as a Recommendation on 1 October 1996 — a stable, reviewed web graphics reference. The same generation of text appeared as RFC 2083 (PNG Specification Version 1.0, Informational) via the IETF. IANA registered the image/png media type. Later, ISO/IEC 15948 published PNG as an international functional specification (the widely cited 2004 edition), keeping the “pronounced ping” note in the abstract. In plain language: what started as a Usenet fire drill became a formal web and ISO-family standard.

Browser reality lagged the paperwork. Native PNG appeared unevenly across Netscape and Microsoft Internet Explorer through the late 1990s. Full alpha compositing was a particular sore spot on Windows IE for years: early versions handled PNG poorly or only with crude binary transparency, and web developers swapped AlphaImageLoader hacks until Internet Explorer 7 improved per-pixel alpha. The format was ready; the dominant browser caught up slowly — one reason GIF banners and loops lingered.

Alpha transparency vs GIF, and why a normal PNG does not animate

GIF89a can mark one palette index transparent. PNG can do that style of simple transparency and, more importantly, a full alpha channel for partial opacity. Soft drop shadows, antialiased logos, and overlays that must sit on unknown backgrounds are the everyday win. That is also why “save as PNG” became the reflex for UI exports and why re-saving a photo as PNG rarely beats JPG on size.

Animation was consciously deferred. Core PNG is one picture. MNG (Multiple-image Network Graphics) grew from the same community as a multi-image / animation sibling — powerful, complex, and never a universal browser default. APNG (Animated PNG) later added animation-related chunks to a PNG stream so older decoders still show a still frame while newer ones play frames. Treat those as side stories: the .png files filling screenshot folders in 2026 are almost always stills.

Why .png files still sit on disks in 2026

In 2026, .png legacy files (and brand-new ones) still appear wherever sharpness and transparency matter: OS screenshots, app store assets, slide decks, scanned diagrams, game UI packs, design handoffs, and “Download image” saves from the web. Cameras still prefer JPEG or RAW for photos; editors still reach for PNG when the pixels must not smear. The format won the still-image half of the GIF replacement bet so thoroughly that people forget it began as an emergency committee on Usenet.

Legacy File Converter lists Portable Network Graphics / .png among supported images on Supported File Types and converts .png locally on the user’s own Windows PC to JPG, PDF, and other listed targets — nothing is uploaded. This article stays the history — not a how-to for writing chunks by hand. For neighboring image-format lessons, see History Lesson: What Is a .gif File?, History Lesson: What Is a .bmp File?, and History Lesson: What Is a .jfif File?.