While blogger was down most of the morning, I recalled a recent conversation about the arXiv's policies. Since not all of the readers around here are registered arXiv users it occurred to me that you might not know it's not possible to withdraw a paper from the arXiv. However, I have good evidence that most scientists are indeed human and occasionally make mistakes - and not in all cases realizing a problem is identical to solving it. The only way to deal with that on the arXiv is to replace the paper with an empty update, or with a dummy file. The previous version(s) of the submission however, remain available on the arXiv.
This happens more often than one might think. I just did an abstract search for 'withdrawn' which produced the warning 'Your query resulted in too many hits, only 1000 hits are being displayed.'.
I don't particularly like this. Nobody is eager to admit mistakes, and replacing a paper with the comment 'withdrawn' means it will appear in the next listing under 'replacements' with a clearly visible stigma. I think that probably some people who have realized a problem with an argumentation prefer to just leave the paper as it is, and hope nobody ever mentions it again. The result however is that weak or wrong ideas remain on the arxiv, publicly available. Yes, it's called a pre-print server, but nobody is required to add a comment saying: This paper has been rejected by 6 different journals because I've abused infinities (see previous post) and I've eventually understood why.
That is to say, for this reason I would think the arXiv quality might improve if it was possible to withdraw a paper - completely - from the listing.
However, on the other hand I am afraid it might encourage users to submit premature works, knowing that they 'could' withdraw if it turns out to be bullshit. Maybe public humiliation is the better tool to guarantee quality? And I guess I would miss some of the stage fright that comes with pushing the 'submit' button (and don't yet relax).
So, I am kind of undecided on that issue, what do you think?