Re: IceBreak as an alternative to Apache?

Den 01/01/11 17.36, Joe Pluta skrev:

A von Neumann machine is one where the program code is modifiable. The
program code for the i is essentially not modifiable, hence non-von
Neumann. The primary upside is that program code can be shared among
multiple jobs, since it can't be modified.

Ah, that's why. To my understanding a von Neumann machine is one where programs and data are stored in memory without consideration if some parts of memory are write protected.

There's no opinion there, so I don't get your "must think of something
else" remark. Are you inferring that I consider the midrange better
than other operating systems because they are non-won Nuemann? I do
think the midrange is better than any other operating system, but the
non von-Neumann nature is only one part of that. And to be pedantic
about it, it's not really "non von-Neumann" since the program code can
be updated, but only by the operating system not by application
programs. So calling it non-von Neumann is more of a convenience term.

I am not being facetious or anything like it. As both OS/400 and AIX run on the same physical hardware now I could not understand your non-von Neumann comment, and asked to clarify.

And just to expand on that, whether or not read-only code is "common" is
not really the point. The point I was making is that the IBM midrange
is built around this concept and takes advantage of it in many ways.
That's why entire enterprises with hundreds of users could be (and in
some cases still can be!) run on machines whose total memory is measured
in megabytes, while a typical Windows user needs 4GB just for their
desktop machine.

I do not agree with this conclusion. The single largest reason for large amounts of memory being needed for a Windows machine is the graphics module which the IBM i doesn't have. Just see how much more the Aero capability for Vista/7 requires in terms of computer power (but it is nice).

I would rather say that the problem with non-server Windows and Linux is the missing pools, allowing one process to gobble up all memory available and use all cpu available, so the solution that has arisen is virtualization which is extremely inefficient but solves the problem.

I spent some time with Solaris which has very advanced capabilities for pooling resources giving similar results to what we see on the i.

So, yes, other operating systems have shared code. But like so many
things, the i has had it for a much longer time. The IBM i - we have
technology before technology is cool<smile>.

I agree that the i appears to be very efficient in how things work.