On Friday, 3 October 2014 at 11:31:07 UTC, eles wrote:
The former attempt stability (because all packages are tested together, along with their interactions), while the latter attempt cutting-edge software (you update software as it gets produced).
This generally true but not entirely true. Rolling release model also implies testing of package inter-operation but any guarantees only apply to versions that match specific repository snapshot - most problems arise from trying to update some of packages but not all. At least this is the case for Arch.