I was only using physics as an example. There are journals that I'd love to read regularly - Leonardo, Computer Music Journal, some of the art photography journals - but I don't, because they're  

1. Not available locally
   
2. Too damn expensive
   

or
3. Both.

And Phys Rev and the rest still matter to journalists, if maybe not so much to scientists, because the content is easier to filter through if you're looking for a story. Arxiv.org is a sprawling free for all in comparison. You might find a story in there, if you're lucky. But it'll probably take you half a day of sifting through abstracts.

One problem with limited public access to information is that often you don't know that you're interested in a publication or book until you fall over it by accident.

Google may be a kludge, but at least you can search for things easily and have a decent chance of finding something relevant.

Not so with paper, where it's not only expensive, it's also inaccessible in the more practical sense of being invisible until you discover it. And your chances of discovering it depend on where you live and how rich you are.