I largely agree that journals charge far too much for subscriptions but they do provide value added. Latex is great for physics and math, but provides little help to biologists. Frankly, after writing grants, doing the work, analyzing it, writing it up, and defending it at conferences, I feel I don't have a lot of time left over to play with margins and get the typesetting and hyperlinked references all working. The layout work actually is valuable. Yes, new tech makes it easier, but there's still the research to do.
Additionally, some journals have staff that help with the review process. Peer review is done by people busy with other things who often miss a lot, espeically well executed fraud. Many of the biology-related publishers perform text and image analysis of submitted articles to look for evidence of fraud. They find duplications, square edges where square edges are never found (introduced through deletions), etc. Not EVERY journal falls into ALL of the stereotypes, and Elsevier is by far the worst offender.
I also find it funny when people blast open access journals for having page charges to authors as if this is a new affront. Virtually all journals (at least in the biochem/biology space) have HUGE page charges and often charge hundreds of dollars extra for each color figure. A lot of color ISN'T used to save money. When the Public Library of Science opened in 2003 they got blasted because they had a flat $1500 publishing charge and then it was free open access from there. That charge was less than half that charged by other journals for just the base price.
Publishing WELL includes editorials, perspective, handling fraud and retractions, etc., and keeping the legacy data available in supplements available to modern computers. I suppose this COULD be done by a volunteer army by it's important enough to pay to have it done well. These are the archives of our knowledge. This may look cheap and easy to the IT crowd but other disciplines don't fall so easily into having 1 server at MIT and some volunteers. It doesn't and shouldn't be as expensive and bound up in copyright as it is (PLoS lets me keep the copyright and it's so nice not to have to ask for permission to use my own figures) but there is probably a happy middle ground as is already been explored by more and more open access journals.