I was under the impression that the arXiv was started in August of 1991. For example, navigating to the hep-th category and pulling up a listing by dates will list the first paper as hep-th/9108001 by Horne and Horowitz. Indeed wikipedia tells us that
The arXiv was originally developed by Paul Ginsparg and started in 1991 as a repository for preprints in physics and later expanded to include astronomy, mathematics, computer science, nonlinear science, quantitative biology and, most recently, statistics.
However…with the arXiv api one can perform searches for papers by the date they were last updated. Here, for example, is how to search for arXiv articles prior to the one listed above http://export.arxiv.org/api/query?search_query=lastUpdatedDate:[000012310900+TO+199108150000]&max_results=100&sortBy=lastUpdatedDate. Whah? Lots of hits, including many papers by Knuth 🙂 Note that these papers do not have arXiv numbers below the one listed above…but they have date stamps that are prior to the August 1991 date! By this account the first updated paper is physics/9403001—a, uh, paper of some note—that was updated on Fri, 25 April 1986 at 15:39:49 GMT!
The moral? Well there isn’t one, but those who play with the arXiv api may need to consider this date oddity.
I think many papers have the wrong date because they use the LaTex today command and the arXiv re-compiles papers to pdf when it’s not in its cache.
Weird, I thought it could be data corruption, but physics/9403001 has obviously been backdated by some arXiv administrator. I bet they have changed their policy. They didn’t backdate http://arxiv.org/abs/1003.2133 to 1929. 🙂
Do you know why so many arXiv papers have the wrong date on the first page of the pdf file?