Everybody seems to worry about downvotes, as if the only purpose of the voting system is to just get more votes. I think that its purpose should be to give an idea of the overall usefulness of a node, not just to convert the system in a competition to get more votes. Maybe some people consistently downvotes some writers, or use votebots; other will consistently upvote those same writers and the use votebots may be either way. Thus, I don't think the effect of those "anomalies" is going to essentially affect the results of the votation on truly good and truly bad nodes.

On the other way, the reasons to up- or down-vote a node may be fuzzy; you may feel that something is wrong (in the contents or in the form) or that the node is wonderful, but you may be not able to actually express way. Also, you may not exactelly remember the exact reasons for a vote without careful reading of the whole thread, which could consume a lot of time, which I (at least) am not wishing to use; perlmonks reading already needs more time that what I have.

Last, I've got the feeling that some people are thinking that they will have the opportunity to challenge the downvotes received, thus upping its reputation. This seems unfair. If challenging is to be implemented, both up and down votes should be challenged, randomly if possible, potentially reducing the reputation. Moreover, it should have such a cost as to make it an option to be taken only in extreme cases. As I think this would be used by people wishing to increase the reputation of a node, such a price could be a number of XP, plus a downvote for each unanswered reply (or something like that). I would be much more in favour of a system giving feedback on the good and bad points of a node (in the line of having several ++ and -- boxes). Moreover, I feel that the anonimity of voting is a good part of the system; removing it would possibly cause some people to stop voting, just to avoid the hassle, but the more votes are overly cast, the more accurate are the results.