On Tue, 16 May 2000, you wrote:
> Derek Martin says:
> And as the article you ponted to earlier suggests, an organization can
> write perfect, bug-free code. They simply need to make the commitment to
> do it. I agree wholeheartedly that the design process is the key. If
> your process is bad, your software will be too, except perhaps completely
> by accident.
>
> Well, actually, on an MS system, you can't. It's not possible to
> write any software at all without calling lower-level software, at
> least at the system-call level, and usually at the run-time library
> level. And since you can't know what that does (because you can't see
> the source code), you can't rely on your understanding of the lower
> levels being correct. While it's possible to design tests of the OS
> and libraries, the number of possible paths is so large that it would
> take millenia to run them on even the fastest machines. So without
> access to the lower-level source code, all other code's behavior is
> inherently unpredictable.
>
> This is essentially the same argument that security analysts often
> use: If you want your computer to do only what you tell it, with no
> surprises, you must have access to the software and hardware specs at
> all levels down to the very lowest IC logic. If you don't have all
> the details on something that you call, you can't know its behavior,
> and you can't predict the behavior of anything that that uses it.
>
> Software on an Open Source platform is knowable and reliable in
> principle, because you can get at the code for all the lower levels.>
Sort of true in practice, possibly not in principle.
See http://www.acm.org/classics/sep95/
While this makes even open software suspect, it certainly make me glad it
exists.
> If it's on undocumented hardware, then you do have the same problem,
> of course, at a level below the software. But software on a secret
> platform like MS systems can't be made reliable even in principle,
> due to the unknowable behavior of the OS and the libraries.
--
Standard is better than better. If your web page cares what browser I'm using
it's broken.
[EMAIL PROTECTED]
**********************************************************
To unsubscribe from this list, send mail to
[EMAIL PROTECTED] with the following text in the
*body* (*not* the subject line) of the letter:
unsubscribe gnhlug
**********************************************************
