Biophysical, developmental and systems-biology considerations enable deeper understanding why cancer is life threatening despite intensive research. Here we use two metaphors. Both conceive the cell genome and the encoded molecular system as an interacting gene regulatory network (GRN). According to Waddington's epigenetic (quasi-potential)-landscape, an instrumental tool in ontogenetics, individual interaction patterns ( = expression profiles) within this GRN represent possible cell states with different stabilities. Network interactions with low stability are represented on peaks. Unstable interactions strive towards regions with higher stability located at lower altitude in valleys termed attractors that correspond to stable cell phenotypes. Cancer cells are seen as GRNs adopting aberrant semi-stable attractor states (cancer attractor). In the second metaphor, Wright's phylogenetic fitness (adaptive) landscape, each genome ( = GRN) is assigned a specific position in the landscape according to its structure and reproductive fitness in the specific environment. High elevation signifies high fitness and low altitude low fitness. Selection ensures that mutant GRNs evolve and move from valleys to peaks. The genetic flexibility is highlighted in the fitness landscape, while non-genetic flexibility is captured in the quasi-potential landscape. These models resolve several inconsistencies that have puzzled cancer researchers, such as the fact that phenotypes generated by non-genetic mechanisms coexist in a single tumor with phenotypes caused by mutations and they mitigate conflicts between cancer theories that claim cancer is caused by mutation (somatic mutation theory) or by disruption of tissue architecture (tissue organization field theory). Nevertheless, spontaneous mutations play key roles in cancer. Remarkable, fundamental natural laws such as the second law of thermodynamics and quantum mechanics state that mutations are inevitable events. The good side of this is that without mutational variability in DNA, evolutionary development would not have occurred, but its bad side is that the occurrence of cancer is essentially inevitable. In summary, both landscapes together fully describe the behavior of cancer under normal and stressful conditions such as chemotherapy. Thus, the landscapes-attractor model fully describes cancer cell behavior and offers new perspectives for future treatment.