Zion How long it takes

Minutes ยท Wait 09 of 36

Building another circuit

Asking for a new path is not the same as asking for the page again. The request goes to software on your own machine, and the pause that follows is the network being rebuilt underneath you.

Zion Market addresses

markz73ffpnbidkldppsj5cjnkfujdsy3rkpi5inqw3ikt4omwsldlqd.onion
zionm3fshzgd5pin3yrc5ueltoykozprcsggrgbl6b3dzbpdtfphv2id.onion

Printed as they were given to this site. Nothing here opens them, so none of this says either string is answering at the moment you read it.

Usually feels like

Minutes

You stay at the keyboard and it finishes there.

Every hop in a new path costs a full trip out and back, which puts the pause above a page load and well below anything you would walk away from.

The wait, in plain words

Something died. A page stopped halfway, a connection dropped, an address that answered a moment ago stopped answering, and the obvious move is to ask for a new circuit. The screen then does very little for a while.

What you asked for is a fresh route through the network, not a fresh copy of a page. Nothing is being fetched during the pause. A path is being chosen and negotiated, and only once it exists can anything be fetched over it.

It sits in this band because it ends at the keyboard and because it is genuinely several round trips long, each of them across machines run by strangers in places nobody consulted you about.

Underneath, while nothing moves

A circuit is built one hop at a time. Your client picks relays, opens a connection to the first, and negotiates keys with it. Then it extends through that relay to a second and negotiates again, then through both to a third. Every extension is a message that travels the whole distance and comes back. The cost is not the arithmetic. The cost is the waiting.

Reaching an address of this kind is more than that again. There is no ordinary destination to dial. Your client fetches a description of where the service is listening, contacts one of those points, and then both sides meet at a third relay chosen for the purpose. Each of those steps is another set of trips, which is why the first answer from an address is slower than anything you are used to and why rebuilding costs more than it feels like it should.

The new path is not chosen to be better than the old one. Selection spreads load and avoids obvious mistakes. It does not know that the relay it just handed you is on a saturated link, or that the machine is a hobby server on a domestic connection. You can leave a path that was merely slow and land on one that is worse.

Stage of the pauseWhat is being waited on
Relays chosenA route is picked from what your client already knows about the network. This part is instant and invisible.
Hops negotiatedKeys are agreed with each relay in turn, and each agreement is a full trip out and back.
Service locatedA description of where the address is listening has to be fetched before anything can be aimed at it.
A meeting arrangedBoth sides are told to meet at a relay picked for that job, and the tunnel is joined there.

That is the trap in pressing for a new circuit repeatedly. Each press abandons a path before you have learned anything about it and buys another unknown. You are not sampling until you find a good one. You are paying the build cost over and over and never letting any path show you what it can do. A path you have actually used is a path you know something about. A path you built and discarded is a path you know nothing about.

Behaving, and not behaving

The difference between a rebuild that is worth doing and one that is not comes down to whether the failure is in the path or in the thing at the end of it.

A path being built

The pause is quiet and short lived, and when it ends the address that failed a moment ago starts answering normally.

A later attempt behaves differently from the earlier one. Different is the whole point. It means the route, not the destination, was the problem.

A path that will not help

Every fresh route lands on the same result, at the same speed, with the same message. Independent routes do not fail identically by accident.

Or the builds themselves stop completing, in which case the trouble is between your machine and the network rather than anywhere near the market.

When every route gives the same answer, the honest reading is that the address is not serving and rebuilding is the wrong tool. That is a wait with no ceiling in it, and it has its own card. When the builds themselves fail, the usual suspects are local: a clock that has drifted far enough for handshakes to be refused, a network that filters the connection before it leaves the building, or software left running long enough to have gone stale.

It is worth separating this from the gate screen. A gate means you reached the service and are being held by it. A failed build means you never got that far. Rebuilding helps the second case and does nothing whatever for the first.

Safe to do meanwhile

The build wants nothing from you, so the discipline is mostly about not interrupting it and not losing track of where you were.

What makes it longer

This wait is short by nature. It gets long mainly by being restarted.

What this card is not. This card is not a guide to configuring anything on your own machine, and it does not tell you which relays you are on.

Everything on this site