The Federation as a Geometric Object
A federation of authority nodes is not a network diagram. It is a geometric object with a topology, a covering, and compatibility conditions. Once you see it that way, the operational rules stop being conventions and start being consequences.
The KTS Global Authority Network is a federation of independent domains, each hosting its own authority claims, connected by a topology worker that maintains the roster and a resolver that composes cross-node evidence. Looked at as a network diagram it is a set of dots and lines. Looked at geometrically it is a covering of a larger claim spaceA covering, in topology, is a family of open sets whose union is the whole space. Every federation node is one open set; every cross-node relationship is an overlap; the whole federation is the space being covered., with each node contributing a local chart.
Why this framing pays
Under the network view, the operational rules — do not modify the anchor node's surface; every worker deploy is a single multipart PUT with all bindings inline; frozen dated snapshots are byte-identical across phases — feel like a list of conventions. They read as engineering discipline, which they are, but that framing misses their necessity.
Under the geometric view, they are what the shape requires. If the anchor node is the origin of the covering, changing its surface changes the origin — every other chart's transition function has to be recomputed. That is not a preference. It is what the object demands.
Similarly: if every worker deploy carries the full binding table inline, no PUT can silently drop an edge. The atomicity of the deploy is the atomicity of the transition update. A PUT-then-PATCH would be two updates the covering cannot see as one.
Frozen snapshots are the atlas's paper trail
The dated-snapshot doctrine — that any block carrying a `built_at`, `topology_version`, or `canon pass` marker is byte-frozen — is the same doctrine a mathematician uses when they publish a paper. The paper is the state of the argument on the date it was written. You do not go back and edit it because the argument moved on. You publish a new paper.
A federation whose snapshots can be edited retroactively has no atlas. It has only a current state, which is exactly the property auditability was invented to preventRetroactive editability of historical state defeats the purpose of audit. A frozen snapshot is the datestamped mathematical equivalent of a signed page in a bound ledger..
The onboarding is a covering extension
Adding a new node — the site you are reading this on will itself become one, in a subsequent phase — is a covering extension. A new local chart is introduced. Its transition functions to every neighboring chart must be defined. The atlas gains a piece. The compatibility conditions have to hold on every overlap where the new chart meets an existing one.
The eight-stage onboarding runbook is not eight arbitrary steps. It is the ordered set of conditions a covering extension must satisfy for the resulting object to still be a manifold.
The larger point
Most infrastructure discussions optimize for local properties: this endpoint is fast, this database is consistent, this deployment is atomic. A geometric framing shifts attention to the global one: the pieces fit. Wherever any two touch, they agree.
You cannot get that property by tuning individual services. You get it by treating the whole system as a single object with a shape, and asking, at every step, whether the shape is still coherent.
That is what the KTS federation is trying to be. That is what any serious federated authority network has to become.