Loading
Knocking on the endpoint…
We ask it what it charges, read its catalogue entry, apply the policy, and then ask it again to see whether the price held. That is two round trips to somebody else’s server, so a few seconds is normal and a slow answer is a fact about them rather than about this page.
Walks a whole purchase in the order a buyer asks the questions, stopping at the first that objects. Nothing is signed and nothing is spent. The payment check asks the narrower question, about a payment you have already decided on.
Over the endpoints WE have knocked on, which is our reach and not the catalogue. A disagreement is between two statements by the SAME party, their listing and their endpoint, so it is theirs to explain and never ours to resolve; it is not evidence of fraud and a seller may simply have moved their payee and not relisted. `comparable` is the denominator: an endpoint that names no payee, or whose listing names none, cannot disagree and is not counted either way.
Which endpoints, and whose, needs a plan. Any single endpoint you name is free: search for it above and its own reading is on the row.
Nothing here says stop, and 2 of the 5 questions could not be answered at all: price-is-readable, listing-agrees. That is our reach, not a finding against the seller, and it must not be read as approval.
One knock has failed, on 2026-10-27, and 1 more are needed before we would call it down. A single failure is a reason to ask again, not a finding.
One mark is one knock, not a span of time. We knock a working endpoint about once a day and back off on one that has been down a long time, so a gap here is where we did not look.
Asked just now. We knock at most once a minute per endpoint, shared by everybody, because our own user agent promises not to be a load.
The half that settles those is a real payment from your own wallet, which is not built yet. When it is, the money goes to the seller and never through us.