What Is Browser- and Proxy-Caching?

The Internet is slow. Therefore the less information you need to send through it, the better.

Imagine trudging to the library every single time you wanted to refer to a favorite magazine article. Of course what you end up doing instead is keeping a photocopy of the article on hand (say, in your desk drawer). Whenever you need to refer to it, you pull out your photocopy--a lot faster than trudging to the library!

The same mechanism is used to "speed up" the Internet experience. Your browser keeps "photocopies" of pages (and graphics/banners) that it's seen recently. When you use the "Back Arrow" on your browser, it doesn't request another copy of the page/banner--it just uses the photocopy it already has. When the desk drawer fills up, or if you haven't looked at a photocopy for a while, it's automatically purged. This is your browser cache.

Big ISPs, corporations, and on-line services (like AOL) take this concept one step further... Their networks are fast. So what they do is create a huge desk drawer that is shared by all users who are accessing the Internet through their speedy network. This is a caching Proxy server.

When your browser requests a page (or advertising banner) from such a large ISP or on-line service, their Proxy server first checks its huge desk drawer to see if already has a photocopy from someone else who has recently requested the same item. If so, the Proxy server sends this "photocopy" straight to the browser without going out to the (relatively slow) Internet to ask the originating site for another copy. Otherwise, it goes and gets a photocopy from the originating site and stores it in the shared desk drawer for future reference. A popular site will get some requests from Proxy servers, but a lot less than if there were no proxy-cache.

As David Gray [VP Marketing, Off-Road.Com] noted in his postings to I-Advertising, server logs do not accurately reflect the number of eyeballs. That's because the server has no way of knowing how many photocopies are in circulation. The server only knows about (and therefore logs) requests for originals. David estimates that his real numbers are 2- to 3-times the amount that his traditional logs show.