nightmare

I mostly - certainly not always - am guided by simulations of stable linear systems where I can trust the component models and the sim software. I don't trust future-state simulations of unstable or chaotic systems, especially if I don't understand the component behavior, the forcings, or the initial states.

Simulation mostly helps me to think.

And of course, being in business to sell stuff, my simulations are rapidly, often concurrently, verified by experiment, which also guides future expectations of simulations.

Is that unreasonable?

The ultimate secure computer system will have absolute hardware protections. Programmers can't be trusted here.

Oh. Oh I see. you thought I was talking about you, specifically.

Well I can't really help that if you see it that way.

It's been proved that there's no such thing as an "ultimate secure computer" there's no way to ensure that your compiler isn't compromised in some way, too. Or that your compiler's compiler wasn't compromised, or that the compiler that you use to compile the tool you use to check to see if your compiler is compromised, wasn't compromised.

Or that the hardware that you use to compile the software that you use to design the hardware for the ultimate secure computer didn't itself inject a vulnerability into the design software that then compromises your new hardware.

And so forth...

I believe that absolute hardware protection is possible. But people don't even use the protection mechanisms that are provided. The structure of c sure doesn't help.

Address space layout randomization is hilarious, the moral equivalent of hiding under the bed.

There's a little MINIX machine running inside every Intel processor sold

- who knows exactly what it does. It's known to have the capability to flash the BIOS even when the machine is powered down (but still plugged in/available battery) and can run self-modifying code.

No, that's why C++ is a great idea even for small embedded applications. You can write insecure code as easily as you can in C. But it's easier to write secure code with it as it makes it easier to write code that enforces some set of common-sense generally agreed upon restrictions.

e.g. you may never directly write to a raw storage array of fixed size without bounds-checking, or allow some quantity measured in positive integers to ever have something that's not a positive integer assigned to it. And it can enforce stuff like that with no runtime resource overhead.

There are other languages that do the same out of the box, but the runtime overhead tends to be higher making them inappropriate for embedded work.

Or rather, no more overhead than building in the same assurances manually in C.

LOL The lady doth protest too much, methinks

It worked for Corazon Amurao.

Apart from the initial bias most of the interesting behaviour in Spice comes from its solution of non-linear component models.

No, but it is unreasonable to decry one sort of simulation because what it predicts is inconvenient when you also rely on another simulation.

Even with an entirely separate code and data address space strict Harvard architecture so that data can never get executed there are still ways to subvert an operating system. It is just a bit harder to do.

That isn't to say that stuff could not be done better. OS/2 was very much technically superior to Windows when it was launched but IBM made such an awful hash of marketing it with PS/2 MCA hardware lock-in that apart from in a handful of niche applications it sank without trace.

Segmented modified Harvard architectures can go a long way to defending against most of the common software failings. Unfortunately flat memory models of interspersed code and data have become the norm and some nasty go faster compromises made to aid gaming speeds on PC video drivers.

I am inclined to the view that hardware safety critical interlocks should never depend on software working and I much prefer it if there is a very visible physical interlock that prevents someone from carelessly firing a big laser or lighting a plasma when I am inside the beam path.

I once returned to a piece of kit to find that the only process still running was the small one that pumped the dead man's handle.

Life gets very difficult for a machine like the 9900 series CPU that suddenly finds its registers (including the program counter) are in ROM!

It is if you don't mind having computers that can only do one job.

Your problem is always that at some point you have to load the program code in from external storage as data and then flip a bit to allow it to execute. Controlling that executable transition is the key.

Sadly the likes of Windows have far too much code executing with the highest level of privileges for very minor speed gain and insanely high vulnerability risk of buffer overrun attacks.

C is unfortunate, but we are kind of stuck with it now. Attempts to improve it have made the syntax more complicated but left in the tendency for cryptic obfuscated code that can easily go wrong.

The ability to coerce an integer into a pointer to anything you like is an intrinsic weakness (avid practitioners consider it a strength).

It makes life harder for the aggressors, but they have come up with clever ways to allocate multiple do nowt threads with synchronised release to defeat even the rather more well crafted Apple OS.

Handing programs memory pointers in such a way that the process will be killed if it ever attempts to write outside array bounds is one effective defence but until recently there were overheads on it.

Security has been so often sacrificed for go faster stripes and the cache line attacks demonstrate just how tricky these things can be.

I believe it has been hacked already.

are

as what amounts to fiddling... running some spice sims, breadboarding some thing to try an idea, swapping parts to see what happens.

changed?

Yes - computers are different to other stuff. A computer that old is pretty much useless, a hifi amp from 1959 can (occasionally) be excellent.

NT

But you can still do a huge lot better than MS Windows

NT

tware

ly

ng as what amounts to fiddling... running some spice sims, breadboarding so mething to try an idea, swapping parts to see what happens.

nt changed?

ty much useless, a hifi amp from 1959 can (occasionally) be excellent.

The only reason why old computers are "useless" is because there is somethi ng else that is better and cheaper. It's simply not worth continuing to us e the old machine. But there is nothing inherently worse about an old comp uter. It's not like they go bad in any way other than how everything goes bad, by wearing out.

On the other hand, an old amplifier does have inherent shortcomings in the same way as old PCs. Newer gear is smaller, lower power and can be made wi th features older gear can only dream of. So no one other than collectors have any reason to use old amps. Better stuff is available.

You can do a lot better than McDonalds too, even in the regime of fast food but at some point someone realized that brand recognition and market penetration/availability were more important concepts than the intrinsic quality of the product.

"quantity has a quality all its own" or that is to say 40,000 restaurants nationwide and billions and billions served can't be "wrong."

Microsoft and McDonalds are either thought of as failures of late-stage capitalism or rousing success stories, depending on how one looks at things. They sell a mediocre product at a price a bit too high for what's on offer and they'll always be there for you. What's not to like? It seems that over the years and decades investors at least have sure liked it a lot..

If both companies were really that terrible at what they do nobody would use them, people would find a way, any way, to do something different en masse.

And if they put way more effort into making their product exceptional they'd have no choice but to charge more. More than most Americans could afford on the regular. Someone would move in immediately to undercut them.

These companies are cutting edge, in the field of walking the razors edge of good-enough engineering.

For the tasks the kiddos were attempting to use the Mac Quadra-class machines for at the time (mid 1990s) at ART COLLEGE e.g. standard definition video editing they were sort of useless when they were new, too!

You set up the edits and and effects you wanted to apply, let the pizza box crank overnight, and hope to God it hadn't crashed or locked up in the morning.

eir

rs

etty much useless, a hifi amp from 1959 can (occasionally) be excellent.

hing else that is better and cheaper. It's simply not worth continuing to use the old machine. But there is nothing inherently worse about an old co mputer. It's not like they go bad in any way other than how everything goe s bad, by wearing out.

e same way as old PCs. Newer gear is smaller, lower power and can be made with features older gear can only dream of. So no one other than collector s have any reason to use old amps. Better stuff is available.

The difference is the rate of improvement. Excellent amps existed in 59, am ong much mediocre & horrible stuff, and the best hasn't got a lot better si nce (the average sure has!). Computers OTOH are the main area of fast techn ological progress.

NT