The reason it works is that your peers read your work, and either refute or cite you. Remember -- everybody knows everybody in a subfield, or at least everybody's advisor. It helps when there is math involved which is either correct or not. If somebody random submits something to arxiv, people know they don't know him or his grad school and that is thus probably a crackpot and so don't bother to read it (except sometimes someone from nowhere gets a result, which would NEVER be seen in the closed peer review process).
EDIT: Also you attach your name to a finished product, not an unattributed ongoing wiki page. Your reputation is everything, so you win or lose based on the articles.
One of the problems in social science academia is that 98% of the articles are completely worthless, and they only serve to get a professor promotion based on an index of citation times journal prestige. The smaller state colleges require their junior faculty to "publish", but they don't have enough to say to get into real journals and the publishing houses fill the need with "Journal of Rural Sociology" etc.
It's true that arxiv works well. Do remember, though, that even arXiv does not allow posting by anyone. If you're not affiliated with a major research institution, you need to be vouched for by current members.
IANA Physicist, but I can't imagine any physics journals feel threatened by arxiv. When someone wins a Nobel prize for work that was published exclusively through arxiv, then you might have a point.
I Am A Physicist, and researchers do use Arxiv quite often: pre-publication papers, or something that is written up and can iterated (unlike journals, Arxiv can store multiple version of the same paper). It is not peer reviewed but good for having feedback on things you are doing.
Of course, publications from the "big guns" go straight to the big journals, and their peer review is more like rubber-stamp if you have the name on your author list.
So the only difference is for the bureaucrats, who want you to show them "proper" papers to be able to move forward, the people doing the actual research only care whether your experiment and explanation is solid or not, can be published anywhere.
(And I'm totally for open journals, even more, use CC-BY on it...)
The reason it works is that your peers read your work, and either refute or cite you. Remember -- everybody knows everybody in a subfield, or at least everybody's advisor. It helps when there is math involved which is either correct or not. If somebody random submits something to arxiv, people know they don't know him or his grad school and that is thus probably a crackpot and so don't bother to read it (except sometimes someone from nowhere gets a result, which would NEVER be seen in the closed peer review process).
EDIT: Also you attach your name to a finished product, not an unattributed ongoing wiki page. Your reputation is everything, so you win or lose based on the articles.
One of the problems in social science academia is that 98% of the articles are completely worthless, and they only serve to get a professor promotion based on an index of citation times journal prestige. The smaller state colleges require their junior faculty to "publish", but they don't have enough to say to get into real journals and the publishing houses fill the need with "Journal of Rural Sociology" etc.