Commercial (and non-commercial) science publishing has evolved as a solution to a number of problems in the market for research results. It has reduced transaction costs by bringing together authors and readers, which is just the simple advantage of market intermediaries. It has delivered added value to readers by filtering out bad work.
It has added value to authors by delivering signals of high quality work. It has added value by sorting, relieving readers from the necessity to identify relevant work in some field of interest. And it has contributed to the value of published work by delivering guidance from reviewers to authors.
But technological changes already have and will continue to erase the value of these services. These services can be provided in much better quality and at much lower costs via open access science networks like SSRN.
All we need to make this work are some simple technical improvements and a few new but simple modes of peer interaction. My conjecture is that commercial science publishing will not survive for more than a couple of years.