NNSquad - Network Neutrality Squad
[ NNSquad ] Thoughts on Google's Voice/Images/Instant Pages search announcements today
Thoughts on Google's Voice/Images/Instant Pages search announcements today http://j.mp/ikGJLJ - - - As per requests, some brief thoughts on Google's announcements today -- related to voice search, search by image, and "instant" page loading. Knocking down barriers to knowledge: http://j.mp/mCAYc5 (Official Google Blog) Voice Search in Google Chrome: Great stuff. I don't have access to the official version yet, but I've been using various similar approaches via browser extensions for some time. Given the high quality of Google's back-end voice recognition system, I expect this to be extremely useful in broad deployment. Search by Image: This appears to be essentially a desktop deployment of "Google Goggles" with additional input mechanisms. Goggles isn't perfect, but can be very useful, and is always fun. I've been asked specifically about facial recognition in this context. Given Google's prior statements on this issue, and the way Goggles works, I would not expect to see Search by Image to be the launching platform for facial recognition capabilities at this time. Instant Pages: This expansion on Instant Search prerenders search results pages that Google anticipates you are especially likely to select, so that the pages can appear immediately (currently in the dev versions of the Chrome browser). A number of persons have asked me to compare the current Google "instant pages" approach with earlier matters surrounding prefetching of pages from search results -- issues that have come up in the past. The primary controversies at that time were bandwidth usage, possible distortion of server statistics (for pages prefetched but not actually viewed), and implications for forensic analysis of user systems and server logs by law enforcement (again, did an actual person view a page, or was it only prefetched?) There are some significant differences in Google's new approach (actually based on a Chromium development path that has been in progress for quite some time): Chrome Prerendering: http://j.mp/lOPLOP (Chromium Projects) For example, Google's prerendering approach will reportedly download all resources associated with a page, not only (for example) the main HTML as in some older prefetch implementations. Obviously, since this could have significant bandwidth implications, it is important that Google's "relevance" decisions regarding when it makes sense to prerender a page be as accurate as possible. Also, the current development approach provides an API that servers can use to determine if a page is in view vs. only prerendered: Visibility API: http://j.mp/kxVcFM (W3C) I do not currently know if it is possible to determine whether any given GET in a server log was the result of an actual user click vs. a prerendering prefetch, at least without invoking such an API mechanism. Google says that they are taking a conservative approach to prerendering decisions, only invoking this feature when there is high confidence that a particular choice is likely to be clicked. This should yield considerably fewer "wasted" page downloads compared with past "unintelligent" prefetch mechanisms implemented by some browsers. Perhaps even more importantly, over the years it appears that law enforcement has become considerably more attuned to the fact that there are various ways for Web page text and images to appear in user PC caches (and in related server logs) without the user having actually purposely chosen to view those materials. So many of the forensically-related risks associated with prefetch/prerendering mechanisms would seem to be of considerably less concern than in the past. The broader deployment of such systems will likely serve to further emphasize this reality over time. --Lauren-- Lauren Weinstein (lauren@vortex.com): http://www.vortex.com/lauren Co-Founder: People For Internet Responsibility: http://www.pfir.org Founder: - Network Neutrality Squad: http://www.nnsquad.org - Global Coalition for Transparent Internet Performance: http://www.gctip.org - PRIVACY Forum: http://www.vortex.com Member: ACM Committee on Computers and Public Policy Blog: http://lauren.vortex.com Twitter: https://twitter.com/laurenweinstein Google Buzz: http://j.mp/laurenbuzz Tel: +1 (818) 225-2800 / Skype: vortex.com