The Archivist's First Unindexed Folio: On the Unseen Structure of the First Web Image
In the summer of 1992, a small team at CERN, led by Tim Berners-Lee, was building a system they called the World Wide Web. The browser, a line-mode terminal application, could request and display text. But a colleague, Silvano de Gennaro, who worked in the computing division, had a different kind of file. It was a photograph, a scanned image of a pop band called Les Horribles Cernettes, founded by CERN employees. He wanted it online.
The problem was one of pure information architecture. The web, in its pristine, early state, had no provision for images. It was a web of words, of linked documents. To include the photo, Berners-Lee and a young intern, Jean-François Groff, had to invent a new tag. They didn’t call it an `img` tag; they called it an ‘icon’ tag, a term borrowed from the graphical user interfaces of the time. The syntax was simple, almost naive: `<icon name="image-file">`. It pointed to a file, and the browser, if it could, would display it inline. It was the first act of embedding, the first time a document’s composition was altered by an external resource.
Today, we speak of Core Web Vitals and Largest Contentful Paint, obsessing over the milliseconds it takes for a hero image to resolve. We fret over layout shifts as text reflows to make space for a tardy JPEG. But the very first image ever served on the web introduced all these performance and stability questions in their most embryonic form. That first request was, by any modern measure, brutally unoptimized. The image format was a GIF, chosen not for its compression efficiency for photographs, but because it was there. There was no `loading="lazy"`, no `srcset`, no `sizes` attribute. The render was entirely blocking; the document would halt, wait for the image data, and then continue.
We can imagine the layout instability, primitive though it may have been. The line-mode browser, rendering row by row, would have encountered this new `icon` element and had to decide what to do. Did it leave a blank space of a predetermined size? Or did it simply render the alt text, ‘IMAGE’, and then have the picture erupt into that space later, a sudden and disruptive arrival? This was the first Cumulative Layout Shift, a historical jank moment witnessed by perhaps a dozen people.
This act of embedding was a trade-off, a conscious decision to sacrifice the pure, predictable flow of text for the richer, more volatile world of mixed media. De Gennaro and Berners-Lee weren’t thinking about bounce rates or conversion metrics. They were thinking about sharing a moment of joy, a picture of their friends. They prioritized the human experience of connection over the technical purity of the document. In that single, unoptimized GIF of a comedy band, they set the stage for every performance dilemma we face today, reminding us that at the heart of every technical constraint is a fundamentally human impulse: to share what we see.
Notes & further reading
A few pages I came back to while writing this: