A Rule You Can Apply Without Arguing

"A gauge may use a keyless public source listed in its row below. Nothing that needs a new key, a login, or scraping-against-ToS enters the roster."

Restating a rule when it stops fitting

The product started with a hard budget: no new API keys. Three providers, that is the whole allowance. It is a good rule — every key is a credential to provision, rotate, secure, and eventually leak.

Then the gauge spec ran into a fact: index-level valuation data is not fully reachable through those three, and some of it lives in keyless public statistical sources — a published CSV, a set of daily figures, the same class of thing the macro provider itself is.

What the spec did next is the interesting part. It did not quietly make an exception, and it did not hold the original rule and ship a worse product. It restated the rule, in a section explicitly headed as restating the provider budget honestly.

Why the new form is stronger

The new rule has four properties the vague version lacked, and each one does work.

It names the real concern. The worry was never "external data", it was credentials and access agreements. A keyless public dataset carries neither, so it was never the thing the rule was protecting against.

It draws a checkable line. Does this need a key, a login, or scraping past a gate? Three yes-or-no questions. Nobody has to arbitrate a judgment call about whether a source is "official enough".

It is per-gauge, not global. Each admitted source is listed in its own gauge's row. A source is not blanket-approved for the product; it is approved for the specific measurement that justified it.

It keeps provenance mandatory. Every stored value carries its source string regardless. Loosening which sources are allowed does not loosen the requirement to record which one produced each number.

When a rule blocks something it was never meant to block, restate it — do not carve an exception. An exception leaves the wrong rule standing and adds a special case, and the next similar situation reopens the whole argument. A restatement fixes the rule at its root, and the fact that it needed restating is itself worth recording, because the original wording will otherwise be quoted back at you.

The gate that stayed closed

The strongest evidence that this is a real rule rather than a rationalization is that it refused something. When sector aggregates for one market were investigated, one path required a session, one sat behind a bot wall, and both were closed by citing this spec's own rule — no scraping past a gate.

A loosening that only ever admits things is not a rule, it is a preference discovered after the fact. This one loosened in a specific direction, held in another, and the record shows both.

Give every admitted source a resilience posture at admission time. Each keyless source enters with the same terms: cache the last-good value, show a staleness indicator, and never let it become a hard dependency for serving. A public dataset with no contract behind it will eventually move, rename, or go quiet — planning for that at admission is much cheaper than discovering it on the morning it happens.

Restating a rule when it stops fitting

Why the new form is stronger

The gate that stayed closed

Code

External links

Exercise

Progress

Comments 0