Before the Image Was Whole

When raster data is delivered in row order, a decoder can paint the portion already available. This makes progress concrete: the boundary between image and blank space moves down the page. Yet the method distributes meaning unevenly. The top may contain background while the lower region contains the object. The viewer knows exactly how much area has arrived and may still know almost nothing about what the picture depicts.

That mismatch made loading itself a small narrative. Portraits assembled forehead before expression; product photographs disclosed backdrop before object; diagrams exposed titles before conclusions. Designers could crop with the sequence in mind, but often the experience was accidental. The file's storage order became the visitor's viewing order because the network was slow enough to let one be perceived as the other.

The PNG specification defines Adam7 interlacing as seven passes through an 8-by-8 repeating pattern. Early passes contain widely spaced pixels; later passes fill the missing positions. The approach does not create information that has not arrived. It selects a transmission order in which a coarse representation spans the image before complete local detail becomes available.

This changes the visitor's earliest question from “how far down has it loaded?” to “what is this?” Recognition can precede resolution. That is valuable because attention can make a decision before the transfer completes. If the image is irrelevant, leaving early saves time. If it matters, the rough preview makes continued waiting intelligible. Interlacing respects the possibility that a person may not grant the file unlimited attention.

The GIF89a specification describes a four-pass interlaced row order. Progressive JPEG, often grouped with these memories, uses different scan-based encoding. Adam7 PNG uses seven spatial passes. The visible family resemblance—a rough or incomplete image refined over time—should not erase their technical differences. Historical interpretation becomes unreliable when a familiar visual effect is allowed to stand in for every mechanism that could produce it.

The distinctions also clarify what an interface reconstruction can claim. A museum animation may preserve the perceptual contrast between sequential rows and a refining whole without pretending to be a bit-exact decoder. Naming the source mechanisms makes that compression honest. The object of the exhibit is not a benchmark of format efficiency. It is the design choice to rearrange arrival so that meaning can precede completeness.

High-speed networks did not end staged imagery. Responsive image selection, dominant-colour fields, low-quality previews and lazy loading all manage when visual information enters a page. Some reduce transferred data; others stabilise layout or make delay feel continuous. A blurred preview is not Adam7, but it inherits the editorial premise that an approximate whole can be more useful than an exact fragment.

This premise has limits. A placeholder can imply content that changes when the final image arrives. Aggressive loading effects can add distracting motion. Images that matter to understanding need text alternatives that do not depend on visual refinement at all. Designing arrival is therefore not a licence to aestheticise slowness. It is a responsibility to decide which meaning must be available before polish, and to ensure that meaning survives when animation or imagery is unavailable.