Glossary

Ambiguity bound
The longest time a paying agent can be left not knowing whether its payment went through. The question asked of each rail is whether its documentation bounds that window, and where the bound lives, for example an expiry field inside the signed payment.
Authorization primitive
The step in which a rail checks a payment, or asks for one, before anything is delivered. The DIM-02 measurement times two such steps separately and never combines them: payment validation and challenge issuance.

Index

Capability record
The set of documented facts this site holds about one rail. Each record is served as a signed credential, one per rail, listed at capabilities/index.json.
Challenge issuance
Asking a client to pay before a service is delivered. In DIM-02 this is primitive B: the emission of a payment challenge, either an HTTP 402 response or a minted invoice.
Confidence interval
A range that shows how precise a reported figure is. On the DIM-02 page the 95 percent interval is for the median, derived after the fact from the calibration parameters; it is a descriptive addition, not a pre-registered quantity.
Conformance check
A test, run on a rail's test deployment as Everyday AI Ltd runs it, of whether a bound stated in the rail's own documentation holds when the check introduces a fault, for example a credential presented a second time, or a payment presented after it has expired. The conformance report runs each check that applies to a rail for a fixed number of valid repetitions on that rail and gives the result as a count; a descriptive check tests no bound and counts only whether an event was observed.
Count
How the conformance report gives the result of each check that tests a bound on a rail: k of K, where K is the number of valid repetitions on one test deployment and k is the number in which the documented bound held, with every VOID and NOT EXERCISED attempt listed by cause. A count is not an estimate of how often this outcome occurs on any network, test or production, and no rate, probability or interval is derived from it.

Index

Deterministic finality
A payment becomes final at one defined moment, after which it cannot be reverted. The methodology names deterministic close as one of the distinct finality mechanisms the rails use.
DID document
A published file that names the key this site signs with. It is served at .well-known/did.json, and a verifier reads the public key from it to check a credential's signature.
Documentation profile
A record of what each rail's own specification or operator documentation says about it, with no figure produced by making a payment. Each cell names the source it rests on and the date that source was read.

Index

Error catalogue
A published list of the errors a paying client can receive. The question asked of each rail is whether its specification or its operator publishes one, and at what address.

Index

Finality
The point at which a payment can no longer be reversed, so the party paid can rely on the money. The methodology defines settlement-finality as the time from submitting a payment to the moment the transfer is irreversible on the rail, at the level that rail's own documentation says a builder may rely on.

Index

HELD and VIOLATED
The two outcomes of a valid repetition of a conformance check that tests a bound: HELD when the documented bound held, VIOLATED when it did not. What counts as each was fixed for every check in the pre-registration, after the declared exploratory runs and before the confirmatory run; a state that the pre-registered table does not list is scored VIOLATED but reported apart as unclassified, never as a divergence from the documentation.

Index

Idempotency
A retried payment does not move money twice. The question asked of each rail is whether resubmitting the identical credential, or signing a fresh one while the first is still valid, can move value twice, and which documented mechanism prevents it.

Index

Local-complete
A rail that can make its decision without a network call on each payment. In DIM-02 a local-complete rail does no per-call network round trip in steady state, and its figures are comparable only with other local-complete figures.
Lognormal parameters
Two numbers that describe the typical value and the spread of a set of timings skewed towards slow outliers. DIM-02 records them per rail from calibration and uses them only to derive the descriptive interval shown beside each median.

Index

Network-dependent
A rail that must make a network call for each payment. In DIM-02 a network-dependent rail makes one per-call round trip, and its figures are not comparable with local-complete figures.
NOT EXERCISED
An attempt in which the check did not test the bound as registered, for example because a refusal's recorded text was not among the reason entries the pre-registration lists for the mechanism under test. It is listed by cause, is not counted toward K, and is replaced by a further attempt, up to twice K attempts in all.

Index

Observer privacy exposure
What a party outside a payment can see about it. The record states, per the documentation, the payer identifier, the amount, any memo or metadata, and whether the payment can be linked on a public ledger.
Offline verifiability
Whether a paying agent can later establish what happened to its payment without asking anyone. The record states whether the documentation defines such an object, for example a signed receipt, who signs it, and the limits the documentation states on that check.
Outcome knowledge
How a paying agent learns whether its payment succeeded. Mediated means another party tells it; trustless means it can check for itself; mediated, then trustless means it is told first and can check on chain afterwards.

Index

Payment validation
Checking a payment authorization and deciding to accept or decline it. In DIM-02 this is primitive A: a payment authorization presented to a verifier, up to the accept or decline decision.
Pre-registration
Publishing a method and its analysis plan in a fixed, dated form, so they cannot quietly change later. Each method here was frozen and externally anchored after its pilot and calibration runs and before any confirmatory run or any published result.
Probabilistic finality
A payment becomes harder to reverse as more blocks are built on top of it, so certainty grows over time rather than arriving at one moment. The methodology names probabilistic rooting as one of the distinct finality mechanisms the rails use.
Provenance label
A tag on each cell saying where its content comes from. The label is one of three, documentation-derived, measured or void, and no documented cell on this site carries the label measured.

Index

Rate limit
A cap on how often a service can be called. The record states only whether the operator's documentation declares such limits on the path a payment uses: a declaration, never a measurement.
Restitution
Getting value back after a payment, or stopping one before it settles. The record lists each kind the documentation describes (a refund, a dispute, or the cancellation of an authorization not yet settled), with who starts it and who bears its fee.

Index

Signed credential
A record whose contents are signed, so anyone can check that it has not been altered and who issued it. Each capability record is a W3C Verifiable Credential signed with ES256 over P-256, checkable offline against the published key.
Specification pin
The exact documents, versions and dates that define a rail's payment flow. Each record names them so a reader knows which text was read; it is metadata, not a score.

Index

Tier-1
Facts that are measured. A dimension measured by a method that was frozen and anchored before publication, shown without ordering by result: timings with confidence intervals, conformance checks as counts.
Tier-2
Facts that are documented. What each rail states about itself in its specification or its operator's documentation, every cell naming its source and the date it was read.
Tier-3
Directions that are declared. Plans stated on the roadmap, none of them measured.

Index

VIOLATED
See HELD and VIOLATED.
VOID
An attempt whose outcome cannot be decided, for one of the causes the pre-registration lists, for example a read needed to decide it failed, or a network's software version changed during it. It is listed by cause, is not counted toward K, and is replaced by a further attempt, up to twice K attempts in all.

Index

For agents and software

The same facts as data, at these paths.

  • index.json the machine index of every served surface