We address the problem of optimal size-selective exploitation in an age-structured fish population model by systematically examining how density- and size-dependency in growth, mortality and fecundity affect optimal harvesting patterns when judged against a set of common fisheries objectives. Five key insights are derived. First, while minimum-length limits often maximize the biomass yield, exploitation using harvest slots (i.e., regulations that protect both immature and very large individuals) can generate within 95% of maximum yield, and harvest slots generally maximize the number of fish that are harvested. Second, density-dependence in growth and size-dependent mortality predict more liberal optimal size-limits than those derived under assumptions of no density and size-dependence. Third, strong density-dependence in growth maximizes the production of trophy fish in the catch only when some modest harvest is introduced; the same holds for numbers harvested, when the stock-recruitment function follows the Ricker-type. Fourth, the inclusion of size-dependent maternal effects on fecundity or egg viability has only limited effects on optimal size limits, unless the increase in fecundity with mass (hyperallometry) is very large. Extreme hyperallometry in fecundity also shifts the optimal size-limit for biomass yield from the traditional minimum-length limit to a harvest slot. Fifth, harvest slots generally provide the best compromises among multiple objectives. We conclude that harvest slots, or more generally dome-shaped selectivity to harvest, can outperform the standard minimum-length selectivity. The exact configuration of optimal size limits crucially depends on objectives, local fishing pressure, the stock-recruitment function and the density and size-dependency of growth, mortality and fecundity.