Happy 10th Birthday Yahoo!

That would mean Yahoo was a year old when I first visited them. Our company was still called Proximus (before the Vicinity name change) and had an office on North Mathilda in Sunnyvale, just around the corner from Yahoo. I remember walking over to visit them and discussing maps (we powered Yahoo and most other portals back then) along with the upcoming driving directions and yellow pages.

Later, a few of us met with their engineers to discuss our business locator, with the possibility of adding our spatial search technology to some of their products. We used Oracle databases along with a number of patented methods of optimizing for really fast location searching. One of their engineers, Foo (not sure on the spelling), was adamant that the only way to get the performance needed for Yahoo was to use flat files.

Yahoo has always been my ideal model for how to make a high performance, high traffic site. Everything about their homepage is geared towards reducing bandwidth. Look at the source code for their homepage: one and two character urls, all spaces and extra markup removed. On top of that most of the site is rendered into flat files without any dynamic content. "Dynamic" pages typically meant they just rendered the file more often.

It was, and still is, a great method. If you have an extra space in a page, just a single space, that means an extra byte gets transferred by the applications, servers, and network each time someone requests the page. Not an issue with a thousand users a day or even an hour, but serving that extra character 25 million times per day means 25 megabytes of extra work/time/cost.

Still, that's not bad, we download files that big all of the time. Now go look at the source for CNN's homepage. 47K of HTML. 2,735 extra spaces (not even counting all of the other waste you could remove). Almost three extra gigabytes transferred a day with only a million hits. 25 million hits would cost 68 gigabytes. Just to transfer meaningless information.

I suppose I should apologize to Pedraum who had to put up with my "every space counts" tirades on a regular basis. ":^)

Long story short, we didn't cave in and rewrite our entire application suite to use flat files. But I will never forget Foo's flat file finder.