Hacker News Reader

Comment by mrob | Hacker News Reader

TempleOS has no process isolation, so it's completely unaffected by Spectre and Meltdown. I'm starting to think there's a lot of wisdom in its design. Complexity of modern computer systems keeps increasing but the value I get from that complexity isn't increasing nearly as fast. Spectre and Meltdown are just recent examples of some of the cost of that complexity. Do we really need an OS too complex for any one person to completely understand?

View on HN

That's like saying "I don't have any doors in my house, just open frames. So I'm not affected by the weaknesses in locks."

Locks are just propaganda anyway.

May I suggest a social construct ?

Yeah but TempleOS also lacks networking so the house is also in interstellar space completely isolated from burglars trying to steal your stuff.

It's a very interesting design, I recommend to atleast play a bit with TempleOS.

If lack of doors was standard, everybody would see the open frames, and culture would adapt to them. Weaknesses in computer security are much harder to spot. Spectre has been possible for decades without mainstream attention. It's this lack of understandability that I think is dangerous. Many people will assume systems are secure just because they don't see an obvious way to exploit them, and because the complexity has grown far beyond what will fit into a human brain there will undoubtedly be non-obvious flaws. Is weak security really better than no security? The "no security" model really means "socially enforced security", which is playing to human strengths. It's much easier to understand than figuring out the full consequences of speculative execution.

I might believe that stance if you can think of any human scenario where there is value and nobody tries to take advantage of it.

We can’t have nice things/utipia as long as some people are evil.

Which could be fine, if we have a society without theft or the house is in a gated community etc. I think that was the point of the grandparent: when do we really need it?

This is not the first OS without kernel/userspace isolation. LISP machines for example used virtual memory but not process isolation.

In fact it was a top advertised feature that you could get your hands on any piece of data and code running on the system at any time.

We generally speak about the first OS with that isolation; of course early OSs were without it.

It's a bit similar to Windows 9x -- anything can do anything.

It might not be suspectible to Meltdown and Spectre per se, but the lack of security that grants that property still means a lack of security.

Windows 9x has process isolation, but has no concept of permissions thus any process can do anything that is provided by the OS API (which includes reading and writing memory of other processes, injecting code into kernel and such things). This means that it is very unlikely for misbehaving process to compromise the integrity of OS itself or other processes (modulo the fact that such process can confuse some other OS component, eg. window manager, such that the system becomes essentially unusable).

You raise an interesting point!

Maybe everything should have access to everything, and then you keep your _real_ secrets, like encryption keys, on dedicated hardware that locks away the key to the outside world and only does signing. Like a Trezor bitcoin wallet or secure enclave on iOS.

It follows the same design ideas of Smalltalk, Lisp Machines, Mesa/Cedar, Lillith, Oberon, JavaOS, ....

The idea being that the isolation can be achieved via the language type system, instead of extra hardware.

People tend to think of kernel/userspace isolation as if the kernel is that piece of rock-solid code and userspace is that which gets you pagefaults.

If only instead of pagefaults you would get thrown into debugger...

>wisdom in its design

Design choices in TempleOS are basically made by rolling dice.

Any program containing some 50-100 KLOC or more is too complex for any one person to completely understand. (Even if that person wrote the whole thing: he or she will have to refer to the documentation, and puzzle over code written months or years ago to figure out what is going on.)

It also has no TCP/IP stack from what I remember.

> Do we really need an OS too complex for any one person to completely understand?

I think the answer to this is yes. But at the same time, I think we're kind of messing up because children today don't have the same kind of systems a lot of us grew up with -- ones that boot instantly into programming consoles with easy ways to make sounds and draw graphics. Terry's vision to create a modern system like that is deeply good, I think. Personally I imagine something more like booting into a console that understood Forth and/or Lisp, and had what we now consider to be reasonable resolution and number of colors... but Terry is building his dream, not mine. Also, God provided a specification to Terry.

The console of today's youth is in the browsers. Any kid can open the developer's console and start typing JavaScript, the results are displayed immediately on the screen. Considering that JavaScript has it's roots in Scheme, it is a good language for kids to experiment on.

For learning lower level constructs maybe Webassembly will help the future generations. The interested kids can also purchase raspberry pi s and dive into a world of low-level programming.

Saying that JS has roots in Scheme is like saying English has roots in Chinese.

From http://speakingjs.com/es5/ch04.html

"In 1995, Netscape hired Brendan Eich with the promise of letting him implement Scheme (a Lisp dialect) in the browser."

And he has stated several times that he was disappointed that it wasn't scheme he was going to implement.

JS is good at being JS, but it has never been close to being scheme.

Would be, if English had been designed by people fluent in Mandarin. If you know Scheme and imagine a curly C-style alternative syntax for it, you basically end up with early Javascript.

> If you know Scheme and imagine a curly C-style alternative syntax for it, you basically end up with early Javascript.

Except without guaranteed TCO, macros, homoiconicity, continuations...you know, the things that make Scheme unique

Javascript is a terrible language for children to start with.

Javascript is a terrible language.

On the contrary I think it's ideal. It can adopt any paradigm you like to learn; imperative, OO, functional...

Just open the console and start typing. Results are immediate. No need for linking libraries or setting-up a virtualenv or whatever.

Reply on news.ycombinator.com