As part of my daily job, we're constantly balancing business value with level of effort when planning which features to include in an upcoming iteration. After reviewing an iteration's work with a client, follow-on conversations usually revolve around adding new features to the product. I was glad to see that Digg sometimes has a different approach:

In the next week or so, we’ll be closing down the podcasts section and folding it into the video section of Digg. We’ll also be retiring the old Digg Spy. Both of these features have become outmoded as Digg has grown and as a result they have a very small number of users (under 1,000) each.

I think that the key to this sort of activity is having a clear reason to remove a feature. In Digg's case, they were keeping an eye on their metrics and were able to justify the removal of both features based on quantifiable evidence.

In the early stages of development, it's not always an option to remove a feature. The best way to make this happen is to launch early and often and let your users decide which features are valuable and which ones aren't.