However; why not turn this on its head, and offer a real-time, or near-real-time, Wikipedia feed service to paid-up subscribers?

Real-time mirrors seem to be a recurring phenomenon. They are a drain on Wikipedia's resources, and hunting them and shooting them down is a continuing battle. The reasoning behind these mirrors appears to be: 1 putting up a Wikipedia mirror with ads will make money... 2 too lazy to set up a proper mirror... 3 instead, set up a script that queries Wikipedia in real time... 4 profit! However; why not turn this on its head, and offer a real-time, or near-real-time, Wikipedia feed service to paid-up subscribers? Currently, Wikipedia's running costs are about $1.2M per year, and this pays for, among other things, serving about 4000 hits per second, that is to say, about 1.26 x 10^11 hits per year, or about $ 10^-5 per hit. (Of course, this is average gross cost; marginal cost will be significantly higher, say $ 10^-4 per hit). Web advertising rates are generally of the order of $1 CPM: that is, $ 10^-3 per hit. If an advertiser manages to get 10,000,000 hits per year, they will make $10,000 in ad revenue, and costs the Wikimedia Foundation around $1000 in leeched server load. What if we were to turn things round, and charge (say) $ 2 x 10^-4 per hit for an official real-time mirror service? (Of course, this would be aggregated in lumps, because it's impossible to bill tiny fractions of a dollar). Now, the economics to the mirror operator is $ 10^-3 - $0.2 x 10^-3 per hit, and they still make 80% of the money they would have before, and don't need to worry about being cut off. However, the economics for the WF are now quite different: instead of losing $ 10^-4 per hit, the Foundation would make $ 2 x 10^-4 income - $ 10^-4 cost per hit, and thus makes $ 1000 gross profit over the course of the year for those 10,000,000 hits, which can be ploughed back into achieving the Foundation's charitable goals (for example, by buying new server kit and bandwidth, or paying for other real-world activities). Note that the users of the real-time mirrors are _not_ being charged for use of the GFDL content, which remains freely available as before; they are being charged for real-time access to WP data, with no need to run a modified copy of MediaWiki in order to run their service. Administration of the scheme could be made automatic, by allowing the existing credit-card interface to be used to for payment, and entering an IP address or addresses to be authorized, an E-mail address for contact, and getting an authorization key mailed back. As a result: * Wikipedia remains ad-free * the WF gets revenue * the advertisers still get to make (slightly less) money, but this time without leeching unauthorized resources. The feed could be provided from the existing software, only with a "null skin" that produced only the rendered page content, thus both slightly reducing the load of producing it (eg. no check for messages, greater possibility for caching), and, at the same time, making the page content easier to re-use, by removing the need to strip the user-interface from around the page contents. With other changes, for example, not checking for red/blue links, serving costs could probably be reduced even further, and quote possibly WF could charge more than $ 2 x 10^-4 per hit. Given the number of mirrors around, setting up this scheme might pay for itself in a month or less. Good idea, or bad idea? -- Neil

* Wikipedia remains ad-free