January 17th, 2005

  • evan

tail calls and security models

(I do get work done, I swear; I was waiting for a compile.)

An interesting thread over at Lambda the Ultimate (aka, where all of my posts come from) about tail-call optimization's interaction with Java's security model: if you naively optimize a function off the stack because of a tail-call there's the potential you're losing the security information (what permission that code has) associated with the function as well. As always, links to some papers are included.

I especially like the comment that says "The answer is so simple and obvious I'm amazed it took so long to find." and links to a 24-page paper full of sentences like "Following Cligner's [1998] work on defining tail-optimized languages via space complexity classes, we first reformulate the λsec semantics as a CESK machine."