Skip to content

Information Locality

Version
v5 · 2026-09-28 · History
Prime #
1380
Domain group
Applied Sciences & Engineering
Origin domain
Computer Science & Software Engineering
Subdomain
Systems Architecture and Information Placement → Computer Science & Software Engineering
Also from
Organizational & Management Science, Law & Governance, Engineering & Design (beyond software), Medicine & Healthcare
Aliases
Epistemic Locality, Information Placement Principle
Related primes
Asymmetry, Information Asymmetry, Context, Mission Command, Locality Of Reference

Core Idea

Information Locality is the placement principle that responsibility for a decision, guarantee, or interpretation belongs at a level that possesses the information needed to discharge it. A less-informed layer may filter, accelerate, recommend, or catch some failures, but it cannot own correctness when decisive context remains elsewhere. The principle is about criterion–information fit, not a blanket preference for physical nearness, decentralization, or low latency.

How would you explain it like I'm…

Whoever Knows Decides

The person who decides something should be the one who knows what's needed to decide it. If you're choosing whether your sandwich needs more peanut butter, you should decide, because you can taste it — not your friend across the room. Your friend can still say "looks a little dry!" but you get the final say.

Decide Where the Info Is

Information locality is the idea that a job or decision should belong to whoever has the information needed to get it right. A coach can set the team's rules for the whole season, but the player on the field decides when to pass, because only they can see the defender right in front of them. Other people can still help — giving advice, doing early checks, or catching some mistakes — but they can't be in charge of getting it right if they can't see the difference between situations that need different answers. What matters isn't who's closest, but who has the right information for that specific decision. And if the information can be sent to someone else quickly enough, the decision can move to them.

Match Decisions to Information

Information locality is the principle that responsibility for a decision, a guarantee, or an interpretation should sit at the level that has the information needed to carry it out. A less-informed level can still help — filtering, speeding things up, recommending, pre-checking, or catching some failures — but it can't truly own correctness if, from where it sits, two situations that need different outcomes look the same. The key isn't physical distance; it's whether the information available matches the criterion being decided. So you always start by naming what must be decided, because different criteria need different information and can belong at different levels of the same system. In computer networks, a router owns the decision about forwarding on its own link, while the application at the end owns whether the whole message arrived correctly. The principle is not absolute: sending, copying, or summarizing information can move where ownership is possible, and cost, speed, privacy, or authority can justify extra checks elsewhere.

 

Information locality is the placement principle that responsibility for a decision, guarantee, or interpretation belongs at a level that has the information required to discharge it. A less-informed level may filter, accelerate, recommend, precheck, or catch some failures, but it cannot own correctness if two states that call for different outcomes are indistinguishable from its position. The decisive relation is not physical distance but criterion-information fit. The analysis begins with a named criterion rather than an architectural slogan, since different criteria require different information and may be owned at different levels of the same system: a router owns a link-local forwarding decision while the application owns whether a complete message met an application-level contract (the reasoning behind the end-to-end argument in network design), and headquarters can own organization-wide resource ceilings while a field unit owns a time-critical choice whose local conditions cannot be transmitted fast enough. The principle is prescriptive but not absolutist. Replication, summarization, sensing, and communication can move relevant information and so shift the feasible ownership boundary, and latency, cost, privacy, consistency, authority, and failure containment can justify auxiliary checks or a knowingly imperfect placement. It says what the information structure permits, not that other design constraints disappear.

Tensions in Practice

Broad Use

  • Computer systems: endpoint-defined guarantees remain at endpoints; application semantics remain where application context is visible.
  • Organizations: operational authority moves toward people who can observe changing local conditions, while central policy supplies boundaries and shared goals.
  • Medicine: a decision requiring bedside signs and patient preferences cannot be fully discharged by a remote rule using summary data alone.
  • Law and governance: responsibility for a judgment belongs where the relevant record, authority, and purpose can be considered together.
  • Engineering and safety: a component-level check can assist but cannot certify a system-level property that depends on interactions it cannot observe.

Clarity

The prime separates the owner of correctness from helpers. An intermediate check may be useful and even highly reliable, yet still lack evidence needed to close the criterion. Conversely, the best-informed location may be expensive or slow, so other levels can perform provisional work without inheriting final responsibility.

Manages Complexity

Placement disputes reduce to five questions: what exact criterion must be discharged, what information is necessary, where that information exists, which candidate owner lacks part of it, and what auxiliary work can safely happen elsewhere. This prevents architectural slogans from substituting for an information map.

Abstract Reasoning

Use the information-deletion test. For each candidate owner, remove every fact it cannot observe and ask whether two states now look identical even though the correct decision differs. If so, that location cannot own the criterion. It may still optimize common cases, but a better-informed level must verify or override it.

Knowledge Transfer

The roles travel unchanged: a criterion, its required information, several candidate owners with unequal views, a placement decision, and a gap that only an informed owner can close. Endpoints and routers can be replaced by field teams and headquarters, clinicians and administrators, or components and system integrators without changing the reasoning.

Example

A network relay can verify that each packet passed a local checksum, but it cannot know whether the complete application message arrived in the required order and matches the application's intent. Relay checks may improve performance, yet the application endpoint must still own the end-to-end guarantee because the decisive information lives there.

Relationships to Other Abstractions

Local relationship map for Information LocalityParents appear above the current abstraction, mutual partners to the right, and children below. Node labels state whether each abstraction is prime or domain-specific; colors identify relation types.Information LocalityPRIMEPrime abstraction: Asymmetry — presupposesAsymmetryPRIMEDomain-specific abstraction: Plan-Execute Gap — presupposesPlan-Execute GapDOMAINDomain-specific abstraction: End-to-End Principle — is a decomposition ofEnd-to-EndPrincipleDOMAINDomain-specific abstraction: Σ-Algebra of τ-past — is a kind ofΣ-Algebraof τ-pastDOMAINDomain-specific abstraction: Locally integrable function — is a kind ofLocally integra…DOMAINDomain-specific abstraction: Monad (nonstandard analysis) — is a kind ofMonad (nonstand…DOMAINDomain-specific abstraction: Thread-local storage — is a kind ofThread-localstorageDOMAIN

Current abstraction Information Locality Prime

Parents (1) — outgoing typed relations

  • Information Locality presupposes Asymmetry Prime

    Information Locality becomes discriminating only when candidate owners differ in access to information material to the criterion they might own.

Children (6) — abstractions linked here by typed relations

  • Σ-Algebra of τ-past Domain-specific is a kind of Information Locality

    The sigma-algebra restricts events to information locally available by a random temporal boundary.

  • Locally integrable function Domain-specific is a kind of Information Locality

    The property is certified on every compact local region while relaxing global behavior.

  • Monad (nonstandard analysis) Domain-specific is a kind of Information Locality

    The monad captures infinitesimal locality around a point.

  • Thread-local storage Domain-specific is a kind of Information Locality

    It remains its own entry because its identity is fixed by the process and threading model, variable declaration and lifetime, per-thread allocation and initialization, lookup mechanism, destructor and thread exit, module loading, memory limits and behavior under fibers or async tasks.

  • Plan-Execute Gap Domain-specific presupposes Information Locality

    Plan-execute gap presupposes that execution holds criterion-relevant constraint information the planning level lacks.

Hierarchy path (1) — routes to 1 parentless root

  • Information Locality → presupposes Asymmetry

Not to Be Confused With

  • Locality of Reference: clustering of accesses in time or space, which supports caching and prediction.
  • Information Asymmetry: a condition of unequal knowledge between parties; Information Locality is a responsibility-placement rule derived from criterion-relevant differences in access.
  • Decentralization: authority may move inward, outward, upward, or downward depending on where the required information exists.
  • Defense in Depth: redundant safeguards at multiple levels can coexist with one level retaining final ownership of a guarantee.
  • Separation of Concerns: modular responsibility is broader; Information Locality selects an owner through access to decisive information.