It's a current debate:

On one hand, many journal charge page charges - in seismology, BSSA, JGR, GRL, all charge hundreds to thousands of dollars per paper, more for color. Some, like Science and Nature, are free to authors but not to subscribers, but not so easy to get papers accepted in. I think GJI is also free, I'm not sure about PaGeoph, but the relative ranking of status sometimes leads people to go to the more costly journals much of the time. Someone has to pay for the cost of publication, which includes some editors, the clerical costs of the associate editors, the cost of printing & mailing, and web preparation and archiving.

On the other hand, many people argue we should just post our work on online archives, with other scientists free to post comments as to their opinion of the work. Governments, Japan is a prime example, are willing to pay costs of the archive, but so far it has not caught on. The biggest pitfall, from the little I've heard, is that few people are willing to take the extra time to make comments after reading papers, so most work remains completely uncommented and thus much harder to assess than work in a traditional journal.

Granting agencies are willing to pay page charges as an extra tax, although of course they value high-impact publications more than uncited or obscure papers.

The system is not perfect, but works for the most part.