> A "Publish early, publish often" strategy could produce more and better results in the long term I think. That (at least the "publish early" part) is already

> A "Publish early, publish often" strategy could produce more and better results in the long term I think. That (at least the "publish early" part) is already the case for a lot of programming language research in things like semantics or program analysis. Published papers often treat a toy language (no arrays, no dynamic memory allocation, etc.) and leave the rest for "future work". Sometimes that future work does materialize, often it doesn't. It's a difficult situation for academia: If the basic system is too hard to extend to a full language, you will end up without final results, and from the point of view of your funders you have just wasted a lot of time and money. If it's too easy to extend to a realistic language, you might build a tool but end up doing "only engineering" without having new breakthroughs to publish, again seen in academia as wasting research funds. For many people in academia, the incentives for building systems really applicable to real-world languages are just not there.