Michael Peel wrote:
Erm... how many people actually know what an interwiki is? I doubt it's a significant number. Combine that with how many people would think about of that particular usage of Special:Search, and I suspect that you're talking very small numbers. Certainly, I've never thought of that in ~ 5 years of using Wikipedia.
John already covered this, but quite a few tools have a need to parse links reliably in an article. One of the most reliable methods (though incredibly hackish) for handling links like this is to pass them through "Special:Search". For example (using curl):
$ curl -Is "http://en.wikipedia.org/wiki/Special:Search?search=wikia:un:UN:N" | grep Location Location: http://www.wikia.com/wiki/c:un:UN:N
$ curl -Is "http://en.wikipedia.org/wiki/Special:Search?search=mw:MediaWiki" | grep Location Location: http://www.mediawiki.org/wiki/MediaWiki
While I can agree that a lot of _people_ likely aren't typing these links like "wikia:un:UN:N" in to the search form, I can see a lot of scripts using this method to parse a page full of links, hackish as it is. Developers have a tendency to use what works, regardless of whether it's hackish. And there's a tendency to let old code rot, so even if an alternative to this system becomes available in the API, many tools will likely continue to use the "Special:Search" hack.
All of these people wouldn't be seeing the page either. So your primary audience would be people searching on Wikipedia for a topic that doesn't currently have an article or a redirect. Given that a another sizable percentage of views comes from search engine results, the pool of actual views you're talking about becomes even smaller.
I don't understand why this is a problem - if Wikipedia doesn't have a page on what they're searching for, then wouldn't they be more likely to click a sponsored link to somewhere else that does?
I'm saying that this is a problem in the sense that the numbers that are being used to make projections are faulty. Advertisers care about people viewing their ads, so people try to take measurements of views (particularly views by humans). In this case, however, it's incredibly likely that the measurements being put forth are horribly skewed. This affects both the projections you can make and the overall conversation about ads and ad revenue that can take place.
The evidence is bolstered by another redirect page ("Special:Random") having so many hits according to the data you linked to. It's not even possible to view that page in any meaningful sense. Put some ads there and I doubt you'd hear many complaints, but you'd be getting millions of "views" each month. ;-)
Special:Random is just plain fun, though, especially when you're getting started with reading Wikipedia. It has a huge amount of popular appeal. As a result, I'm not sure that it's quite comparable to the search function, which is obviously much more orientated at finding a specific page/description...
Err, I think you might have missed the point here. The comparison is a page in the Special namespace that isn't viewed much, even though the stats say it's viewed millions of times a day. Take a look at the link Liam provided (http://wikistics.falsikon.de/2009/wikipedia/en/). It says that in 2009, "Special:Random" received (on average?) 2,385,287 hits per day. The point I was making is that this is completely misleading, as every hit to "Special:Random" redirects to an article. _Anything_ you place on "Special:Random" would never be seen (it's just outputting a 302), even if it's allegedly getting over 2.3 million views per day. (As an aside, the curl trick referenced above also works with "Special:Random".)
A unknown (but likely very sizable) number of hits to "Special:Search" are in the same category: they're not actually viewing "Special:Search", they're just using it to resolve a link or resolve their search box input. That's where the "be careful" warning came from.
Calling "Special:Search" the most popular page (or basing fundraising theories on it) is dangerous and often misleading work.
I'm not convinced of this assertion yet.
Well, you should be.
MZMcBride