Content negotiation helps to represent a resource at the same URL in different ways considering the capabilities of the requesting user agent. It allows user agents to choose between several representations of a resource by giving an appropriate MIME type in the HTTP-Accept-header.
Browsers typically accept MIME types of text/html to get some human readable representation. A semantic web application, though, might prefer to get a representation of the resource in RDF and thus accepts only application/rdf+xml. In BibSonomy we have added this behaviour to URLs representing particular resources - i.e., a single bookmark or publication reference (which may be represented by several posts). You can access this feature by prepending the /url/ and /bibtex/ URLs with the prefix "/uri/", e.g.,
http://www.bibsonomy.org/uri/url/d1bb7b3f6cafafa7b418f9f356ff2e83
or
http://www.bibsonomy.org/uri/bibtex/2b8b87c78e9e27a44aacde0402c642bff
Depending on the Accept-header of your user agent you get a redirect to the appropriate representation of those resources. You can read some more information on this in the BibSonomy help system.