Business Week featured an interesting article about Yahoo's strategy and social software as a global paradigm on the Web. The author (Ben Elgin) address Yahoo's bet: changing the way people find information online by relying on "social search". Although I am a regular user of flickr or del.icio.us (not to mention others) and though I find the 'social search' idea useful/relevant for my interests, I think I can buy some of the author arguments:

It's time consuming: "Most Internet users haven't even heard of Flickr or del.icio.us, let alone spent time sharing photos online or posting bookmarks of their favorite sites. Alexa Internet ranks del.icio.us as the 364,886th most trafficked Web site. Google is ranked third by the researcher. (...) the first major effort involves selecting a circle of friends. That means e- mailing people, inviting them to join a network, and responding to requests from others. After that, the more users interact with content, the more power social search will have. But that could involve more time-consuming online activities, from simple bookmarking to labeling and reviewing Web sites. It's not clear users will make that kind of investment.

Others doubt the wisdom of crowds will offer much of an upgrade over the feats of raw computing power. (...) "The best description of a document is the document itself."

As with all community sites, the benefits grow with the size and activity of the group. That means Yahoo's social-search trial, still in its infancy, could take months or years before reaching its potential. "Social search is not one of these things that will take off overnight," says Forrester Research analyst Charlene Li. "It will take a lot of time to build."

I don't want to play the part pooper but some of those claims are important and it seems that some Web2.0 platforms try to address this by various means: not allowing the 'tag thing' but having au automatic parser of the document (text/pictures...) for example.

Still, it does not mean that it can't work but there are some things to consider. I would add that in some domains it can work; especially when there is a certain density in conversations about a certain topic.