Publish before peer review
February 27, 2014  

Traditional academic publishers are gatekeepers: conferences and journals won’t publish papers before they pass through peer review. This is supposed to ensure quality, but in my experience it rarely improves papers, and mainly serves to delay them.

Any academic will have a handful of horror stories on the topic, but Leslie Lamport has a particularly fine collection. Graduate students or anyone new to academic publishing should take a look, paying particular attention to his recollections on publishing the Paxos algorithm. His comments on another paper provide insight on what to expect when trying to publish interdisciplinary work.

Fortunately, we have other options today, and the era of these mediocre gatekeepers is coming to a close: researchers are voting with their feet. Here are two recent data points.

The bitcoin paper

In November, Ittay Eyal and Emin Gün Sirer published a paper on arxiv claiming that the Bitcoin protocol was flawed. This attracted a lot of attention given the current Bitcoin frenzy; it was widely reported and drew comments from many other researchers.

So far, so good for the authors. And in March, the paper is going to appear in a conference, Financial Cryptography and Data Security 2014; publishing on arxiv first didn’t prevent them from publishing in a traditional, peer-review-gated venue.

The obfuscation paper

Another paper that has gotten a lot of publicity recently is Candidate Indistinguishability Obfuscation and Functional Encryption for All Circuits, by Sanjam Garg, Craig Gentry, Shai Halevi, Mariana Raykova, Amit Sahai, and Brent Waters; here’s an article in Wired, for example. It was first published on the Cryptology ePrint Archive, which, like arxiv, does not require peer review. Later it appeared in FOCS 2013; publishing before peer review, once again, does not prevent traditional publishing.

Notably, the eprint version of the paper was cited in other papers (e.g., this one) before the traditional publication occurred. And it is common in cryptography to cite papers on eprint.

The end is in sight

It’s clear that in computer security, at least, authors can’t afford to wait for their papers to pass through peer review before publishing. There’s no point in waiting—waiting won’t improve your chances, and publishing first won’t hurt them. Furthermore, waiting will hurt you because others could publish the same results first. So it’s over.

I’m not sure how long it will take, but academics aren’t dummies: eventually this is going to spread to all of science. In fact, computer science is a lagard here; physics and math were the trailblazers. Computer science is strangely slow to get the Internet.