Real-time mirrors as a net revenue _source_? - Wikitech-l

9 Apr 2006

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