I was shooting the breeze with Geoff Challen in Zürich before we headed down to HotOS and we chatted about an idea my advisor and I had chatted about a few years earlier. The idea is that there should be some statute of limitations on how long an old systems paper can cause new papers to be rejected because they aren’t novel. The justification is that after some reasonable period of time—say 10, 15 or 20 years—the assumptions of the paper are probably out of date enough that attacking the problem again even in a similar way is probably interesting to the community.
After the conference, Geoff blogged about the idea in better detail and with a few extra ideas that are worth thinking about.
I just wanted to add a few comments. First, the idea (or at least my interpretation of it) isn’t to ignore older work, in fact, not citing the old paper should probably carry the same penalty that it does today. Instead, the idea is that you should be allowed to publish similar work (as long as you cite it) some period later and not have it hurt you the way it does in the current system.
To some extent I think PCs already do this and will look favorably on a paper which makes use of old techniques as long as it provides some new value as well. So, I’m more calling for an increased shift in that direction than anything else.