Part of the issue here is that we're running into a topic I am extremely, unreasonably nerdy about, since the first post-RTM working overflows are the K-T boundary of computer security. Having a working stack overflow on SGI MIPS, a delay-slot architecture with a split I&D cache, would have been a very big deal, and would force me to revise a mental history I've been grooming for a very long time. I believe Aleph One. :)
(We go back too, him and I; I met my wife at a party at his apartment).
Oh, his version of the history of his seminal paper is definitely canonical, no doubt about that - but I definitely remember seeing him hacking on stack exploits in the period leading up to when he released the paper, whether he had things working or was otherwise probing, and maybe there was a draft/discussion or two that we're both not remembering quite right. I remember him snarfing my DEC Alpha for a few quick checks, too .. Halcyon days indeed.
(We go back too, him and I; I met my wife at a party at his apartment).