Right, but in that case you lose completely the crucial gatekeeping function. Already now arxiv.org is full of papers with glaring technical and conceptual mistakes. Imagine what would happen if submissions to arxiv.org where they way one decides who to promote.

At the cost of being a retrograde, I believe that this sort of things cannot be settled simply "the google way" i.e. 1. Count the citations. 2. Make a list. 3. Profit.

You need to "force" people to review their papers, correct mistakes, and in a sense establish what is worth other people time. Journal editors and reviwers screw up all the time, no doubt, but the burden of proof is on those arguing that "drop it on arxiv, let the review count do the scoring" is a better system.

Also, once you drag the taxpayer in, you have to figure out who polices the archive, least it becomes crackpot central: should they be elected? Should they be appointed? Who should pay them? What about conflicts of interest?