[cap-talk] Midori in The Register

[David-Sarah Hopwood]

Ivan Krstić wrote:
> On Aug 5, 2008, at 10:49 AM, Jonathan S. Shapiro wrote:
>> I am not aware that any guarantee stronger than these exists. Have I
>> missed something?
> 
> Presumably that's exactly James' concern: the apparent conflation of  
> actual security with execution correctness (as promoted by managed  
> code). Execution correctness, while a step in the right direction, is  
> several astronomic units away from solving the key desktop security  
> problems we have today.

Well, yes, but what's the problem with promoting the use of managed
code? "Managed code" just means verifiable CIL/MSIL code run by a CLR
implementation [*].

You *always* have to take Microsoft marketing claims with a cellar
of salt, but that's a separate issue from the technical merits or
demerits of CIL or CLR. I'd prefer that we discuss those, rather than
making vague and unsupported sideswipes about "Microsoft's latest big
idea, an idea which I think leads to disaster", as James did.

In any case, I have not seen Microsoft make any specific claim along
the lines of "managed code necessarily implies better security".
At least, nothing beyond quite reasonable statements about benefits
of memory safety and type safety in addressing certain classes of
"programming mistakes that often lead to security holes", as [*]
puts it.


[*] <http://blogs.msdn.com/brada/archive/2004/01/09/48925.aspx>

-- 
David-Sarah Hopwood