Personally, I don't look for a specific set of features when evaluating a service like this. Instead, I first make sure I can take my data with me (via CSV export or the like) and then I look at how the developer works. I get the impression that Time at LT spends as much time listening to his users as writing code. In the year I've been using LT, it's gone from a reasonable way to catalog my books and find like-minded folks to a full-blown community.
In this talk from the Zend PHP conference, Adam Bosworth discusses the importance of online community. His comments directly support Rob's comment. If you're running a service like GuruLib, the feature set is just the bait that gets enough people using the system to build the community that really adds value. The feature set on Flickr supplies a small percentage of overall value to users.
As Adam says in his talk, there's no rules to follow on what it takes to build a service that supports community. There are some examples, however and they provide some pretty good evidence of what works. I don't know that I'm rady to enumerate those here, but it is something I think about from time to time. Any suggestions on what it takes to build community?