Subject: Re: type safety
From: Matthias Felleisen (matthias@ccs.neu.edu)
Date: Mon Feb 04 2002 - 11:16:31 EST
1. The rewriting semantics and the type soundness theorem form a
specification for the compiler. If something other than the behavior
predicted in the theorem shows up, the compiler is wrong.
2. My original message says that we must include "division by 0" and
friends in "going wrong". Type errors as defined by those that the type
system catches are never signaled. In sound languages, such as ML,
Scheme, and Java, they don't happen at run-time. In C/C++, they happen
but are not discovered per se.
-- Matthias
This archive was generated by hypermail 2b28 : Mon Feb 04 2002 - 11:16:32 EST