I don't understand how you can get away with leaving out proofs in a publication. Given the constraints on the length of the paper, I could imagine providing external references to more complete proofs, say in a companion technical report. It's one thing to trust provided proofs, when even then they are error-prone. But how can you believe a theorem stated without even any indication of a proof? That's kind of the point of math.
I'm reading a couple of papers on FreshML and FreshO'Caml, and they're great work. But the ICFP paper from 2003 has a number of unproved theorems, and then they posted an erratum stating that one of their theorems is flawed, but they haven't bothered fixing it because they have a new paper that takes a different approach. This kind of thing makes me generally nervous about the reliability of research and publications.
No comments:
Post a Comment