The title above was a recent (somewhat contentious) MathOverflow question by Igor Pak. While I still think this is a slightly problematic framing, it’s a discussion people seem to truly want to have, so I thought I might as well create a space for it.
Let me start discussion with the biggest concern (I don’t know if it is a downside) I have about the arXiv:
What is the arXiv’s policy on what can be posted and what can’t? I ask this as an honest question (with some trepidation about getting flaming from both sides). On MO we have an FAQ that tries to lay out as clearly as we can what sort of material should be put on MO and what should not. I cannot find an analogous statement from the arXiv about what they will accept and who they will accept it from. I doubt I disagree very much with what they do in practice, but the lack of an easily located statement of what that practice is actually disturbs me a bit. I apologize if such a document is publicly available somewhere on the arXiv website, but I maintain I shouldn’t have to hunt for it.
After reading the discussion below, I’m even more convinced that the idea of “downsides” is so tied up in one’s values and personal experiences that its impossible to come up with a list that makes sense to everyone. For example: one serious worry seems to be that if you put your papers on the arXiv, people will write follow-ups to them before you’ve had a chance to fully process your ideas. I’ve had this (sort-of) happen to me, though from a talk, not the arXiv; I gave a talk about my research program, and about 6 months later got an email from a graduate student saying he and a collaborator had solved one the problems in my talk. And you know what, it was great. He’d found a reference I hadn’t that made it possible to do lots of other stuff he hadn’t thought of, and I got to farm out that part of the research program to his paper. It really was a problem I wish I had a lot more often.