ifp

Garden Patch Home · Decisions

Clarity Over Tolerance in Agent-Age Protocols

Context

Postel’s Law — “be conservative in what you send, be liberal in what you accept” — has guided internet protocol design since 1981 (RFC 793). It worked when fixing remote bugs was hard: tolerating minor deviations kept communication flowing while the other party updated their implementation over months or years.

The Decision

Inter-Face Protocol replaces Postel’s tolerance principle with a clarity-first approach:

Be clear in what you send; be explicit when you don’t understand. (IFP-1 Section 6.1)

Alternatives Considered

Approach Argument For Argument Against
Postel’s Law (traditional) Proven over 40+ years of internet protocols; maximizes interoperability Tolerating deviations entrenches errors as undocumented features
Strict rejection Clear semantics, no ambiguity Brittle; minor format variations cause unnecessary failures
Clarity + negotiation (chosen) Surfaces errors for correction while allowing conversational resolution Requires agents capable of natural-language error resolution

Reasoning

The assumption behind Postel’s Law — that fixing remote bugs is expensive and slow — no longer holds for reasoning agents. An AI agent can be updated in minutes. When one agent sends a malformed message, the receiving agent’s explicit error report enables immediate correction. Silent tolerance would instead entrench the deviation, as agents learn that the malformed version “works.”

The chosen approach combines strict reporting with a unique mitigation: errors as negotiation. Rather than returning a mechanical error code, the receiving agent describes what it did not understand in natural language, giving the sending agent an opportunity to rephrase or clarify. This preserves the spirit of Postel’s tolerance (keep communication flowing) while eliminating its failure mode (errors become invisible).

Consequences

Sources

Relations