Thursday, May 19, 2005
Someone used the new Pixel Group app and drew a cat on it. Only minutes later, the cat was replaced by a close-up of an eye. This is so well-done I almost suspect the Ajax-service has been accessed via automation.
The Gmail IM Logger plugin for Trillian Pro “logs your IM conversations to Gmail"
allowing you to “keep all of your conversations in one place regardless of what computer you used.”
Today, the Google PR team will be hosting a factory tour (“an informative day for journalists and industry analysts,
covering a wide variety of aspects of Google’s products and business.”) A web cast is available.
Update: Notes and paraphrases from the Web Cast
Marissa Mayer, Director of Consumer Web Products (promoted “God knows how many times”, says Google CEO Eric Schmidt) talked about Google’s core – search:
- 70-20-10 rule. 70% for search and ads.
- Google Maps with Satellite images “best mapping application on the web.”
- Rhetoric question: is there a coherent vision, or do people just build what they want?
- Most product announcements made by Google fall in the 20% category.
- The core 70% is not much talked about – it goes by relatively unnoticed by the public.
- You don’t wake up in the morning being happy you don’t have a toothache; similarly, you won’t notice how often Google manages to not screw up with search results.
There are behind the scenes heros at Google – the “70% engineers.” They care for right results, the right rankings, the great user experience.
- Like Quality Czar Ben, who joined in October 1999. He’s an expert on search, and he can tell which is a good and which is a bad result. He works hard on search ranking. He cares for more relevant “onebox” results on top of the search results. There was a crisis of too many “onebox” results – Ben helped solve this. Ben leads the Google search quality engineers.
On a sidenote, Ben is obsessed with socks (“the bottleneck is socks – if I would have more socks, I would do less laundry”, Ben says). He solved this as well by acquiring a large collection of matching socks.
- Bwolen, Crawl tech lead, joined in 1999. He is responsible to find “more and fresher stuff.”
- Matt, the Spam Cowboy & Porn Cookie Guy joined in January 2000. He built safe-search and tries to avoid surprises for the user. He reduced “spurious porn” and spam results. He reaches out to webmasters to educate about best practices. To remove the spurious porn, you first have to find it. (He bribes co-workers to find porn in exchange for his wife’s cookies.)
- The User Interface Design team. “The homepage always looks the same, what do you work on all the time?” people ask. To assure the user interface looks the same! They added more than 50 features (like News headlines, definitions, Q&A) with little change in the UI. Google delivers a clean, uncluttered site.
The Google User says:
- “I want to always have a Google search box”. So Google developed the Google Toolbar.
Director of Search quality, Peter Norvig, gave a talk on Crawling, indexing and serving 101. A user said, “If Google came in a box, I would eat it for breakfast.” Peter wants to explain “the magic.”
The web, vs Google’s copy of the web. Crawling the web by following links and sucking down pages, making sure there’s not too much load on webmaster’s servers, and no duplicates. In the end, 8 billion or so pages are stored. Now you need to figure out what to do with them.
PageRank is one important attribute (“How many other pages point to this page?”). Important links count more in PageRank. It’s an iterative process which repeats the formula several times. After PageRank, the index is organized.
Peter says: We went out to get clip arts of fancy servers. They look really nice. Our servers don’t look so nice – with all the wires coming out.
Q&A session:
- How many servers?
Not giving out the number, says Peter. “A lots and more.” The last confirmed number was 10,000, Marissa adds.
- The Beta issue. Stuff stays in beta for a very long. What’s the philosophy here? (Constant improvements are made anyway.) Marissa says they have a list of features which they make up-front, which they want to have completed before they let the individual products leave beta.
- How often do you do the crawling and indexing? Peter answers. We’ve evolved that over time. In the past, it’d be roughly monthly. Then we noticed, people ask for fast-breaking news results. The update is now continuous. Key sites, like CNN, are crawled every 10 minutes or so. Other on a daily, or weekly basis. Crawling frequency is proportionate to the frequency of changes taking place on the page.
- Progress and competitors. “We think we have the best results ... we keep trying hard to make it better.” There’s always new challenges producing great results.
>> More posts
Advertisement
This site unofficially covers Google™ and more with some rights reserved. Join our forum!