Mon

Jun 13
2005

Tim O'Reilly

Tim O'Reilly

Unintended Consequences of Google Pre-fetch

Lauren Weinstein's posting today on Dave Farber's IP list argues that Google's recent decision to pre-fetch the top search results for visitors using Mozilla-derived browsers has a number of unintended and (to his mind) undesirable consequences:

  1. It will distort the economics for advertisers on the pre-fetched sites, creating "page views" that may never actually be seen by any user;

  2. It will distort the browser statistics for Firefox and other Mozilla browsers, making them appear more heavily used than they actually are;

  3. More seriously, it exposes users to the cookies for all of the pre-fetched sites, whether or not they were visited.
Lauren goes on to explain how to disable the prefetch behavior in Firefox and other Mozilla browsers, and takes Google to task for not thinking through the privacy implications of their new feature. I'm not as much of a privacy worrier as Lauren, but I'm glad he's bringing attention to the issue. I'd be interested to hear what Google has to say in response.

tags:   | comments: 7   | Sphere It
submit:

 
Previous  |  Next

0 TrackBacks

TrackBack URL for this entry: http://blogs.oreilly.com/cgi-bin/mt/mt-t.cgi/4100

Comments: 7

  Bob Lee [06.13.05 09:30 AM]

To #1 and 2, Firefox sends a prefetch header. It's not perfect, but user experience should be more important to a web site owner than perfect stats. To #3, there's no reason why Firefox can't delay accepting cookies until the user actually visits a site.

  Steve Loughran [06.13.05 09:41 AM]

What worries me that 'browsing prohibited material' is often a terminating offence at the workplace. If the proxy logs say my PC fetched content from a forbidden search, just because a search put it near tthe top, I would have a hard time proving my innocence.

  Bob Lee [06.13.05 10:31 AM]

I just came across the FAQ on the Mozilla web site: http://www.mozilla.org/projects/netlib/Link_Prefetching_FAQ.html. It seems that you can actually tell the browser to make another If-Modified-Since request when the user actually follows the link. So, to #1 and 2, this actually enables you to capture more statistics.

  Phil Ringnalda [06.13.05 10:37 AM]

Of course I know better than to expect care and reason from people who use a proxy log as an HR tool, but: if you can get fired for accessing prohibited HTML without images or CSS or external JavaScript (since prefetching doesn't parse and fetch external resources), chances are you've already been fired for accessing an expired domain that redirects to pr0n, or carries pr0n ads. (And although I guess some text ads aren't in external resources, most are, which mitigates the ad-prefetching problem.)

Before too many people start trumpeting turning off prefetching (which will also turn it off for very useful things like rel="next" links in weblog archives, and which will get blindly added to copy-paste preference tweak lists for years to come), I'd really like to see a single example of a dangerous prefetch. Google isn't doing prefetch for the top result for every search, only for searches where they are quite sure the top result is what you want. So, what innocent keyword are they sure you want pr0n for?

  Trenton Lipscomb [06.13.05 06:43 PM]

First, Google does this for only the first result, not all of them. So, it's not really "top search results" but "the first search result".

Secondly, if you're relying on rows in your access log to determine how many human beings look at your site or see an ad, you're already broken. The IAB has standards for counting ad impressions, page views, etc.

You've got to follow these if you're selling ads in any volume.

  Daniel Brandt [06.15.05 11:11 AM]

What the world needs now is a Firefox plugin that masks AND prefetches all the sponsored links on a Google page. That will make Google less attractive to advertisers, which is the first step toward returning web power to the people. At www.scroogle.org there's already a CSS ad masker for Firefox that works well on Google results pages; it probably could be expanded. Then we can think about looking for AdSense on all web pages, which also might be possible to prefetch.

Mozilla.org should be ashamed. I know they get tens of millions per year from their cozy relationship with Google, but this prefetch thing is going too far.

  Bill de hOra [06.15.05 01:19 PM]

"If the proxy logs say my PC fetched content from a forbidden search, just because a search put it near tthe top, I would have a hard time proving my innocence."

Steve, the referrer can traced back to Google. But you/they could just not use Google. Observation: I'm using search engines in inverse proportion to the number of RSS feeds I'm subscribed to. Search engines are mostly delgated to finding documentation these days.

Post A Comment:

 (please be patient, comments may take awhile to post)






Type the characters you see in the picture above.

RECOMMENDED FOR YOU

RECENT COMMENTS