Re: Tagged memory (was Re: Memory Upgrade)

["Leif Svalgaard" <leif@xxxxxxxxxxxxx>]

> Both Frank and Leif are correct. They just happen to be talking about two
> different things. I say this without having read either of their books, but
> with knowledge of how the system works.
>
> When Frank Soltis and others say that pointers cannot be counterfeited they
> are talking about programs that are created through the machine interface
> (MI) by the trusted translator. All of the compilers on the system must use
> the trusted translator. It will not create a program that will manufacture
> a pointer because the trusted translator enforces the MI and the MI
> includes no mechanism for arbitrarily constructing pointers.

Ed puts it well: as long as the trusted translator produces your
program you secure. This is protection by software. In fact, with
the trusted translator being the only 'producer' of code you don't
need hardware protection at all. Java on the AS/400 runs without
the tagged pointers. So one more time: the protection (or detection)
is SOFTWARE. Taht was my small point.