If I thought I had such a proof (and I expect the proof would be complex, relying on many leading-edge pieces of wildly different branches of mathematics, and therefore comprehensible in the short to medium term to but a few), I would first see if I could develop a practical implementation of the proof for a well-known NP problem, test that out, and see if it worked.<p>If so, I would have that reviewed by trusted colleagues. ("Hey, buddy, I have a P space solution to travelling salesman!" "Get outta town!", "No seriously..."). That would at least demonstrate that the proof works.<p>Next step? Well, there's a bit of the responsible disclosure argument at play: If P=NP and you have a practical implementation of a P-time algorithm for an NP-complete problem, translating that to another will be less work, I should think, than the original proof or the original implementation...<p>...meaning much crypto would break soon after publication. Give the proof and the implementation to 100 well-known and trustworthy mathematicians from around the world, have them agree to your disclosure strategy, then announce what you've got, with their backing, and tell the world that you won't publish for 6 months. Or 12. Whatever.<p>The Fields Medal will wait.<p>That will give the world time to adjust to its new reality.