At least in the popular weblog consciousness, the Barabasi-Albert model of network growth is dominant. BA, for short, essentially claim that popular existing nodes get links from new nodes in proportion to their popularity. Rich get richer, power law, yadda yadda.
Besides skimpy supporting evidence (and no independent corroboration) I've always felt a bit uneasy accepting this model lock, stock, and barrel for weblogs and RSS feeds. Enter Filippo Menczer.
Menczer is a computer science and informatics professor at Indiana University. Among other interesting things (focused crawlers, data mining), he published a paper last year on a Web growth model based purely on local information. See the BA model implicitly assumes each new node has global information. While the model matches empirical data, it doesn't have much basis in reality.
Menczer not only proposes a new model that incorporates semantic distance as a linking factor, he also explores other alternative models for Web growth. In short, there's a number of different ways of explaining those "power laws" besides global popularity and this in fact has consequences for information seeking in such an environment.
Good stuff.