If science is valuable and draws readers who pay subscriptions, then the journals should pay the researchers for their research reports - instead of the 'pay to publish' model that exists today. Journals claim they are the ones providing services to the researchers (in terms of publicity, career advancement, funding opportunities, etc.) - but what about the public's rights to information largely created with public funding?

The reason that publishers don't want to go to the open-access model (which makes sense for a product that is largely taxpayer-financed, i.e. government agencies make their reports available to the public without fees usually) is that universities pay thousands of dollars a year for each journal they subscribe to. (There is archival value for printed paper versions of journals, however.) This becomes particularly important for research into known subjects, e.g. if you want to get the complete history of a certain antibiotic's discovery and production and testing history, at the level of primary research reports (which include methods etc.), it's still very difficult to do that without access to a university library with a comprehensive set of journals, and that's what publishers are trying to protect by going after sites like sci-hub.

The model that Nature and Science want to perpetuate is their 'high impact value', meaning if they accept a paper, it's good for the researcher's career in terms of getting tenure and more funding, and a switch to open-source publishing on arxiv-type platforms might disrupt that stable system. A sticky point is the peer-review system, which is kind of a mess anyway.