Dear Reader,

I dislike advertisements. They tend to be loud, they tend to flatter, and the things they praise rarely asked for it. I am writing one anyway, because the work is finished, and finished work should be allowed to leave the house. The only advertisement I know how to defend is an honest one — so this will be short on adjectives and long on verbs.

What we have built is a measuring instrument. What it does is quickly told.

It finds the distance from which a thing shows its true face. One must keep the right distance from things: too close, and you see only parts; too far, and you see only fog. With people, everyone knows this. With machines, markets and processors, one forgets it — they are not offended when you look at them wrongly.

It hears when something begins to tip. Systems do not collapse out of a clear blue sky. They clear their throat first: they become slow to recover, the way a tired person does. That throat-clearing has usually been sitting in your data for weeks. We read it out loud.

It separates the thing from the lens. Two instruments, two numbers, one and the same system — many a good engineer has despaired over this. The answer is unspectacular: part of the number belongs to the system, part belongs to the lens, and the two can be told apart. This sounds like philosophy. It is bookkeeping. The philosophy comes afterwards, and it is the best part.

And while we are at it: there are measurements that watch, and measurements that join in. We can tell you which of the two you are currently making. Most people do not know.

And it can hold its tongue. This, I am afraid, is the actual sensation. Most analyses deliver a number no matter what — delivering numbers is, after all, their profession. Ours, in such a case, says: there is nothing here to measure. That sounds like little. But consider for a moment how many decisions rest on numbers that exist purely out of politeness.

On the mathematics behind all this I will be brief, since nothing is more suspicious than a long paragraph about one's own seriousness: it is written down, with proofs, as is proper. Part of it is peer-reviewed and in print; the rest is on its way. Wherever we had to assume something, it says so — by name. In this trade, that is unusual enough to be worth mentioning. The method has grown more modest than it first set out to be. In return, it is now proven rather than asserted. Of these two conditions, the second is by far the rarer.

The recipe, by the way, is public. We publish our methods under an open licence (AGPL), and we do so on principle: a measurement you cannot inspect is merely an assertion, and we are in the business of the opposite. You may read the recipe, cook it, and improve it — the licence asks only that you share your improvements in turn, which strikes us as ordinary good manners. Companies that prefer to keep their kitchen doors closed may purchase a commercial licence instead: a way of buying out of the sharing, openly and at a stated price. We find this more decent than pretending to a secret.

What we sell beyond licences is the answer itself: a single page that states where your system will give way first, how much margin remains — or that all is well and you may sleep soundly. The last of these is a calculation too, not a courtesy. Courtesies are the most expensive kind of error in the measuring trade.

And no, publishing the recipe does not worry us. Owning a cookbook has never yet produced a dinner.

If your system ages gently and predictably, you do not need us. We congratulate you — you are in the minority — and wave kindly as you pass.

To everyone else, a proposal: send us a single curve from a system that worries you. You will receive the page. If there is nothing on it, we will say so. That, too, is an answer, and rarely the worst one.

With the reservations of someone writing an advertisement, and the confidence of someone who has measured,

Forgotten Forge