On 6/18/07, Steve Summit scs@eskimo.com wrote:

There's a discussion going on at [[Wikipedia:Reference desk/ Computing#Wikipedia Random Article]] about how the random article feature works. A couple people have made claims that I find surprising, such as that we "take 1000 articles every few minutes and then chose a random one from it", or that the function "only shows articles up to a certain size", or that instead of choosing a random article it now chooses "a random subject (and then a random article in that framework)".

Is anything like this going on? Are we no longer simply taking a a random selection from the entirety of the page table? (Feel free to answer at RD/C, or here.)

Every page has a random number assigned to it in the database. When selecting a random page, a second random number is selected, and the page whose number most closely matches the second number (actually the smallest number greater than the second number) is returned. Basically, the page is simply picked at random, to most intents and purposes. The relevant logic is in SpecialRandompage::selectRandomPageFromDB().

Steve Summit wrote:

There's a discussion going on at [[Wikipedia:Reference desk/ Computing#Wikipedia Random Article]] about how the random article feature works. A couple people have made claims that I find surprising, such as that we "take 1000 articles every few minutes and then chose a random one from it", or that the function "only shows articles up to a certain size", or that instead of choosing a random article it now chooses "a random subject (and then a random article in that framework)".

Is anything like this going on? Are we no longer simply taking a a random selection from the entirety of the page table? (Feel free to answer at RD/C, or here.)

I put an entry in [[Wikipedia:Technical FAQ]] a few months ago answering this question.

-- Tim Starling