our i386 hack to block stack overflows ..

[Theo de Raadt <deraadt_(_at_)_cvs_(_dot_)_openbsd_(_dot_)_org>]

By the way, the hack we have in mind that might let some more of our
"buffer overflow prevention" security improvements work on i386, it
might cost 2-3% performance; that is my rough estimate.  I figure that
propolice itself slowed us down by 1-2%, not that I can tell.

I'm glad I don't put much basis on yammerings on slashdot, or I might
develop a perception that 2-3% is too much slowdown for the security
gained.

In fact, I'm real glad.  We'll try.