Boundary as Interface

Definition

Weinberg argues the everyday metaphor of a "thing" cleanly separable from its surroundings by a boundary is mostly a convenience borrowed from experience of physical space, and it breaks down for most real systems. He recommends replacing "boundary" with "interface" — a two-faced zone that looks both inward and outward at once, since a boundary often connects as much as it separates. Separately, his "Strong Connection Law" holds that what earns a collection the label "system" (rather than an arbitrary grouping he calls a "granfalloon") is that its parts are, on average, more tightly interconnected with each other than with anything outside — "a system is a collection of parts, no one of which can be changed" without affecting the rest.

In the Book

Weinberg works through cases where the clean "skin of a person" model of boundary fails: the layer of air a body constantly exchanges with its surroundings via convection currents, a problem astronauts' zero-gravity capsules had to solve explicitly because convection stops without gravity; and whether human hair is "inside" the body (living tissue) or "outside" it (an excretion, like sweat or fingernails) — the useful answer depends on whether you're studying thermal regulation, parasites, or trace-element excretion. His forest example shows that plants and herbivores, or a Texas saloon with its bar on one side of a state line and its floorshow on the other to dodge liquor and decency laws respectively, resist any sharp inside/outside line. He derives the Strong Connection Law from a practical observation: as science progresses, weakly connected "granfalloons" (his example: people who happen to be from Indiana, versus members of a tight political cell) get decomposed away first, leaving only genuinely tightly coupled collections that still deserve the name "system" — hence "systems, on the average, are more tightly connected than the average."

Why It Matters

This reframes boundary-drawing — a team's edges, a product's scope, an ecosystem's limits, a legal entity's liability — as a design choice with real consequences rather than a fact to be discovered, and it gives a concrete test for whether a proposed grouping is a real system or just a convenient label: do the parts actually constrain each other more than they're constrained by the outside world? It also names a common failure mode — treating a boundary as an impermeable wall — when the more useful move is almost always to study what crosses it.