notes: research/papers/hardman-opaque-identifier-aliases

This data as json

path title type status classification words body
research/papers/hardman-opaque-identifier-aliases Opaque Identifier Aliases summary v0.1 — 2026-10-15: drafted by Claude from the HTML page (version 0.10); unread by Mark; marked partner on Mark's word 2026-10-21, paraphrase only partner 474 # Opaque Identifier Aliases **Abstract.** A 2026 primer by Daniel Hardman on aliases, the private, friendly labels people attach to opaque cryptographic identifiers such as keys, DIDs (decentralized identifiers) and KERI (Key Event Receipt Infrastructure) autonomic identifiers (AIDs). It explains the properties an alias must have, why a shared naming convention (COIA, Conventions for Opaque Identifier Aliases) helps, and why a name that is proved must never be confused with one that is guessed. It is for designers of wallets and identity software. ## Summary The primer presents itself as the companion promised at the end of the author's parable on Zooko's Triangle, explaining the reasons behind the COIA convention rather than restating it. A person acting through many facets, each with its own identifier, ends up with many unreadable strings; the alias keeps that manageable. An alias, older literature's petname, is a private nickname, not an identifier. It is local, not unique, changeable, and must never be parsed for meaning by anyone but its creator. Treating it as an identifier, caching it as stable or reading meaning from a stranger's alias each produce bugs or attack surface. A convention teaches by example: well-formed names lead people to record who the party is, in what role and in what scope, the same facts software needs. The central security point is the difference between proved and guessed. An alias for a remote party asserts a binding that may only have arrived with the identifier; COIA marks such an alias with a leading 0 meaning "unconfirmed" until the parties compare identifiers over a trusted channel. Because people compare long strings poorly, the primer points to entviz, a picture of an identifier that a substituted key would visibly change. Software can usually choose the right identifier from context and should ask only when an act is valid in more than one capacity. Connections are made to facets, not people. Introductions can be automated; the final human judgment that evidence suffices cannot. ## Key terms alias, petname, opaque identifier, autonomic identifier (AID), facet, COIA, proved versus guessed, man in the middle, entviz, Zooko's Triangle ## Claims and their evidence - An alias must be local, non-unique, changeable and never parsed for meaning by others. [informative] - The gap between a proved and a guessed binding is where a man-in-the-middle attacker operates. [informative] - COIA uses a leading 0 to flag an alias whose binding has not been confirmed. [informative] - People are poor at comparing long strings character by character, so comparison should be turned into recognition. [informative] ## Related Conventions for Opaque Identifier Aliases (COIA); Zooko and Houdini: A Flatland Parable; Identity Facets; Amplifying Difference: Perceptual Design and Verification of Human-Centric Entropy Visualizations; Measuring the Glance ## Provenance Drafted by Claude on 2026-10-15 from the HTML page at dhh1128.github.io/papers/oia.html, read in full. Not yet read by Mark.

Links from other tables