What a second address actually buys you

The benefit in one sentence

You gain a next move that does not involve searching. That is the whole of it, and it is worth stating that plainly because everything else attributed to redundancy is either a consequence of that sentence or is not true.

The value is not that three routes are more likely to work than one, although they are. The value is in what you do in the moment when the route in front of you does not work.

Why the next move is the valuable part

Consider the two versions of the same bad moment. In the first, you hold one string, it does not open, and you now have to go and find another. Finding involves a search engine, a forum, a directory, a message from somebody, or a page you half remember. Every one of those is a channel somebody else can put a string into.

In the second version you hold three, one does not open, and you paste the next. The bad moment lasts about eight seconds and never becomes a search. The gap between those two versions is the entire dividend that plurality pays.

The three things it does not buy

• It does not buy availability. Three routes to one service share whatever the service itself depends on. If the service is not running, the number of routes to it is irrelevant.
• It does not buy speed. A quicker route on one attempt tells you about that attempt. The study on one mirror slow and one quick takes this apart properly.
• It does not buy proof. Three strings that look related do not vouch for each other. Whoever produced one could produce three, and the shared head is public. See the shared head.

The shape of the failures redundancy covers

Redundancy helps with failures that are specific to a route. A single onion service can be unreachable while the thing behind it is perfectly healthy, for reasons that have to do with the network path rather than the market. Those are the failures a second address is for.

It does not help with failures that belong to the service itself, and it does not help with failures on your own machine. The section on absence spends its whole length on telling those three cases apart, because the number of quiet addresses is the only evidence a reader has.

What the reader has to do to collect the benefit

Nothing about the benefit is automatic. It is only collected by a reader who is actually holding all three strings at the moment one of them fails. Somebody who saved one address and knows the other two exist somewhere has bought nothing at all.

That is why so much of this site keeps returning to the mundane question of where you keep them. The interesting mechanics are upstream. The value is downstream, and it is collected or lost by whether the text is in reach.

A fair way to size it

A reasonable summary is that redundancy converts one kind of problem into another kind. It converts a search, conducted under time pressure, into a paste. Searching under pressure is where most of the genuinely costly errors on this subject happen, and a paste has no error modes to speak of.

Sized that way, it is a modest mechanism doing one useful job. That is a more accurate description than either the dismissive reading or the version where three addresses somehow make a market robust.

A way to feel the size of the benefit

Think about the last time something you use every day was briefly unreachable. The cost was not the outage. It was the twenty minutes of working out whether it was you, and looking for somewhere that would tell you.

That is the part the second address removes, and it removes it completely rather than shortening it. Either you have a working route within a few seconds or you have learned something real about the situation, which is the subject of the absence section. Neither of those is a search.

What this study is not. It is not a claim about how often anything fails, because this site measures nothing and would have to invent a number to say.