Top Google pagerankers exposed…

Results 1 – 10 of about 25,270,000,000 for -32768. (0.04 seconds)

Enter most any query, including a numeric one, into Google and it will return all the pages which contain that term or are pointed to by links containing that term.

But not long ago, I couldn’t remember whether -32767 or -32768 was the maximum negative int16, so I tried to Google them each and see which had more hits. And I got back… the entire Google index, in PageRank order.

I found that entering any large negative number as a search query had the same result. When changing the query the specific “first page” set, and their order, may shift slightly (or not), most likely as a result of your query being farmed out to one of hundreds (or more) of clusters that are not necessarily synced with one another, but the specific number entered doesn’t seem to greatly affect this result set.

Equally interesting is what happens (or rather, doesn’t) when viewing the cached version of one of these results. In this case, Google’s cache retriever does not attempt to highlight the term on the page, give notice that it only appears elsewhere (e.g. links pointing to that page), or indeed, any acknowledgement that you used a search term at all.

What is that?

A little more playing shows that the same effect applies to any other text preceded by a “-“. My best guess is that Google’s server is interpreting this as an “exclusion” search term (the same way you might search for “bush +kate -george” to get the singer instead of the president) without checking if the query actually includes a positive search term as well. In this case, an exclusion search for some string of text that doesn’t exist on any webpage would simply return the entire index. (And since there is no other measure of relavence, since all pages equally don’t contain the term, it will default to using PageRank to sort for relavence.) Whoops!


One Response to “Top Google pagerankers exposed…”

  1. […] Google are still mad at me for exposing a huge bug in their search some years ago? (In theory, this would make it trivial for someone to determine […]

Leave a Reply