Different holographic dark-energy models are studied from a unifying point of view. We compare models for which the Hubble scale, the future event horizon or a quantity proportional to the Ricci scale are taken as the infrared cutoff length. We demonstrate that the mere definition of the holographic dark-energy density generally implies an interaction with the dark-matter component. We discuss the relation between the equation-of-state parameter and the energy density ratio of both components for each of the choices, as well as the possibility of non-interacting and scaling solutions. Parameter estimations for all three cutoff options are performed with the help of a Bayesian statistical analysis, using data from supernovae type Ia and the history of the Hubble parameter. The $��$CDM model is the clear winner of the analysis. According to the Bayesian Information Criterion ($BIC$), all holographic models should be considered as ruled out, since the difference $��BIC$ to the corresponding $��$CDM value is $> 10$. According to the Akaike Information Criterion ($AIC$), however, we find $��AIC$ $< 2$ for models with Hubble-scale and Ricci-scale cutoffs, indicating, that they may still be competitive. As we show for the example of the Ricci-scale case, also the use of certain priors, reducing the number of free parameters to that of the $��$CDM model, may result in a competitive holographic model.

37 pages, 11 figures, 3 tables, statistical analysis improved, accepted for publication in Phys.Rev.D