Reference
The tests
Sometimes you know what you can look at but not what you are looking for. This page turns the board around and leads with the test.
Read the right hand column first. If one of those is something you can do in the next minute, the pair it belongs to is probably your pair.
| No | Pair | What decides it |
|---|---|---|
| 01 | Down for now, or finished for good | Silence never tells you. Wait for a second signal: a changed page or a changed set |
| 02 | Old address, or one built to be misread | Read the middle of the string. Edges right, middle wrong means it was built to fool you |
| 03 | Same page at two addresses, and only one is theirs | Judge the address in full against a published set. The page itself is not evidence |
| 04 | The redirect and the string it stands in for | Count it. Fifty-six characters plus .onion is an address. Anything less is a promise |
| 05 | Your path through Tor, or the far end | Change only your side. New circuit, then a different onion service you already know |
| 06 | Did the set move, or did your copy slip | Count to fifty-six, then look for 0, 1, 8 or 9. A real address contains none of those |
| 07 | Loading proves a server answered, and nothing more | Name the action you came for, then see whether that completes. The front page is not it |
| 08 | Rate limiting against a password that no longer matches | Stop attempting, wait out the window, then spend one attempt you are sure of |
| 09 | When the puzzle is broken and when the reader is | Reload the challenge without typing and see whether it comes back new and legible |
| 10 | A door you shut yourself against a lock somebody changed | Retrace where you last entered the credential and whether that address came from you |
| 11 | Logged out by a clock against logged out by a decision | Log in once more and read the first screen after it, not the one that threw you out |
| 12 | A code that does not match against a clock that does not | Compare the clock on the code device against a trusted one, down to the seconds |
| 13 | An absence against a single character out of place | Read the name you hold against the record you made at the time, one character at a time |
| 14 | A wait somebody planned against a wait nobody did | Ask whether the wait is a page somebody designed or a request that never finished |
| 15 | A deposit that is slow, or a deposit that is gone | Look the transaction up and read the destination address off the chain record |
| 16 | The address expired, or it was never issued to you | Does your own signed-in deposit record list that address at all |
| 17 | A rule about release, or a decision about you | Can you name the event that releases it, and is it tied to one order |
| 18 | Who took the difference, the operator or the network | Anything missing before broadcast is the operator, anything the chain records is not |
| 19 | A saving on quantity, or a saving on protection | Does the lower price survive an ordinary order through the ordinary flow |
| 20 | Money back, or the end of the argument | Ask whether accepting the amount ends the claim, then read the order state after |
| 21 | Still moving, or already refused | Check the spendable balance: missing means queued, returned means refused |
| 22 | Still settling, or short by a fraction | Compare the amount that arrived at the address with the amount the invoice asked |
| 23 | A date with something behind it, and a date without | Ask what happens on the day after. A window has an answer, an estimate has none |
| 24 | One request moves a clock, the other moves the money | Ask where the balance sits one minute after you press. Held is time, gone is release |
| 25 | Asking for a ruling, or handing over information | Ask what the process ends with. Money moving is a dispute, a note on conduct is a report |
| 26 | A word in a database, and an event in a record | Ask whether any record outside the market says the same thing. If not, it is a claim |
| 27 | Nothing new to report is not the same as bad news | Is there a new entry saying something changed, or only an absence of new entries |
| 28 | A rule that runs by itself, and a date somebody prefers | Ask who can move the date. If the person who named it can move it, it is a preference |
| 29 | Silence and being ignored look identical | Whether the account has done anything at all inside the market since you wrote |
| 30 | Quiet is not the same as gone | Whether the account still appears in the market listings when you look it up fresh |
| 31 | Feedback tells you a mood, evidence tells you a fact | Whether a stranger could look at the same thing and reach the same conclusion |
| 32 | What a signature actually proves | Where your copy of the key came from, and whether you had it before this message |
| 33 | The market only speaks in one place | Whether you navigated to it yourself or it arrived in front of you |
| 34 | An empty history and a claimed history | Whether the continuity is anchored anywhere outside the new account itself |
| 35 | The result is good. Whose key was it? | Compare the fingerprint of the verifying key against one you already held |
| 36 | The same clean result, months apart | Look for a date the signer wrote into the signed text itself |
| 37 | You cannot check an image, however clear it is | Try to select the text. If it will not select and copy, there is nothing to verify |
| 38 | One of these strings was chosen by somebody | Ask which string a person could have typed, then compare the one they could not |
| 39 | Posted where, and written by whom | Ignore where it sits. Ask whether it is signed by a key whose fingerprint you hold |
| 40 | Who granted it, and what were they promising | Ask for a signed line and check it against a fingerprint you already hold |
| 41 | Two lists of addresses, two different guarantees | Check whether the addresses sit inside the signed text or merely beside it |
What makes a test good enough to be here
It has to be available to you. A test that depends on somebody replying, on an operator explaining themselves, or on a status page nobody publishes is not a test, it is a wish. Everything in that column is something a reader can do alone.
It has to land on one side. A test that returns maybe for both halves has not separated anything. Where a test only rules one half out, the page it belongs to says so plainly rather than letting the column overstate it.
It has to be cheap. If a test takes an hour, nobody runs it while annoyed, and a test nobody runs is worth nothing. Almost everything here takes under a minute.
The tests fall into four families
- Compare two strings. Character by character, both directions, no eyeballing the ends. Half the addresses group is this and nothing else.
- Look at a record that is not controlled by the party involved. A payment on a chain, a carrier own tracking, the market own order state.
- Change one variable and repeat. A fresh circuit, a different device, a second address from the set. The value is in changing exactly one thing.
- Ask what would have to be true. The only family with no action in it. Used where nothing is checkable and the answer has to come from reasoning about who could have produced what you are seeing.
A test is not a verdict
Running the test tells you which half of a pair you are in. It does not tell you the situation is fine, and it does not rule out the possibility that you are in a third situation nobody put on the board. Read the page rather than the row.