I'm not an academic, but I think you'll find many academics would be against this. Things would get more complicated because there would obviously need to be a vetting process on who could comment. Things shouldn't be a popularity contest, so I think a reputation system would be counterproductive.
But it's a different type of popularity contest. Commenting on publications is virtually non-existent. It's not that there is no venue of commenting, we do have plenty of academic conferences; but the amount of comments received in the conferences is virtually none. People do ask questions, but seldom give constructive (be it positive or negative) comments.
I don't think this is true in general (though it might be so in the field you are working in). I've received many constructive comments from colleagues, and this has often lead to subsequent collaboration.
Well, I find CS venues (conferences and journals) to be about 50/50 in terms of blind or not. Also, while it will certainly vary subfield to subfield, once you are familiar enough with an area you can pretty confidently guess what group (if not what exact authors) are behind any given paper. I bet it would be pretty easy to cook up a automated paper un-anomymizer based on the body of published work out there already.
In my field of laser physics, there are some really, well known lasers, so if your paper talks about this or that facility and cites this "monumental work by X group" you can easily gather who is the author. Not to mention (yes heaarsay) of at least one person in my group who refereed a paper, harshly criticizing it paper only to have the editor come down and say "accept this paper anyway." Academia has some good parts but it's as political as anything, if not more so.
