Ahh.  When we talk in the e2.0 community about collaboration, we're mostly talking about group collaboration and online collaboration tools-- predominantly wikis.  Andy McAfee recently did an excellent post on this where he explains the benefits of each tool for each purpose.  When a group is looking to collaborate on a deliverable or a group-defined goal or project the best tool is probably a wiki or an otherwise group-editable document platform.  The difference is there is generally a beginning and an end to a collaboration effort.  There could be on-going collaboration efforts, but they are discrete units, usually marked by time or a goal or outcome.  

A wiki is the classic Enterprise 2.0 technology for a core of strongly tied knowledge workers who are collaborating on a deliverable. They can use it to generate documents, to debate their contents and structure, track project status, link to other resources, etc. Google Docs and Spreadsheets, Zoho, and other online office productivity suites are similar to wikis in that they allow egalitarian editing of documents, spreadsheets, and presentations by all group members; they’re just not currently as extensible as a full wiki.

Conversations, on the other hand, are more akin to what we are familiar with in the realm of social networking.  Blogging, microblogging, facebook statuses, feeds and updates alerting us to what our friends are doing.  Conversations are continuous, never-ending.  We are always engaged in the conversation, although we may eat, drink, and sleep intermittently.  Yet these conversations are becoming ever-increasingly global, borderless.  The platforms for conversations are changing all the time.  (I've even used a telephone, and sometimes I speak face to face to humans) ;-)  

On Friendster's demise, it seems to be another victim of Silicon (death) Valley curse, as well as the founders' inability to add features to the product to keep the user community interested.  Although, admittedly, there is a lot of herd-ism to the hot products and platforms on the edge, the tools that have the greatest utility and experience for users will prevail.  Not really so revolutionary, as market acceptance goes.