A study just released by a research group including AltaVista and the I.B.M. Research Center in San Jose, Calif., took a look at the structure of the Web through tracing the links -- "the degrees of separation" between Web pages. They checked the links of 200 million Web pages (about 1/5 of the Web total) and what they were connected to. Out of this they sketched out a structure for the Web which has a kind of " tendrilled spaghetti with a solid core" image, and contains the following components:

There is a strongly linked core of about 28 percent of the Web, something like 56 million nodes -- portals, CNN, Microsoft, search engine links, popular sites (Mudcat!), and so on;

There are about 44 million pages that you go into and don't really come back out of -- commerce sites, and intranets -- these are one way links in, and few links out (the commerce ones presumably to try and trap you into staying there);

There are about 44 million pages that are called "newbies", which are just new sites that are not yet wired into the search engines, or just starting their connectivity;

There are about 44 million pages that are called "tendrils", that have limited connections to anything;

And about another 10 million that don't link to anything, though no one can be sure how many there are because they are not picked up by any search engine -- personal web pages, etc.