This is a wall I have hit with App Review quite a few times

You have to stop viewing the apps from a technical perspective. App Review is _intentionally_ non-technical, as frustrating as that is for us developers. They enforce App Store policy from a non-technical standpoint

I produce an app which can download and run code, and was rejected many times through App Review. I was able to point to many other apps which _technically_ also downloaded and ran code — but the difference, to App Review, was entirely in the appearance (my code was text based, it "looked like code" and not like Scratch's visual programming, or nodes, or whatever)

I disagree with the App Review decision in this instance, but I think they are enforcing policy consistently. It just looks random to those of us with enough technical experience to look past the visuals and to the implementation

We'd be happy to include your experiences when we talk to Apple about 2.5.2 if you'd like to share them with us. (I work on iSH, so you can either use the email on my profile or any of the other ways to contact the iSH team.)

I think JavaScript/Clojurescript is the exception because they control the runtime and JIT: whether or not you buy their explanation, I think the reason they state is that this is to enable sandboxing. This is at least plausible to me: to have effective sandboxing, you need to restrict the APIs that a device can call, which means blocking the ability to generate and run arbitrary machine code on-device.

> to have effective sandboxing, you need to restrict the APIs that a device can call, which means blocking the ability to generate and run arbitrary machine code on-device.

Please, no, this is not how security and sandboxing works at all. Otherwise, the moment you have an arbitrary code vulnerability, you'd get full access to everything.

The way effective sandboxing works is by giving a process a set of capabilities/permissions, what have you, that is enforced by the kernel. Those permissions can be "can I open files", "can I talk to the camera subsystem", etc. Then, even if you somehow manage to generate the right function call at runtime, the call will fail. And to be clear, this is how apple's own security works.

Wouldn’t you have this as part of a defense in depth strategy? Apple’s public vs. private APIs and similar things seem to indicate that at least part of their security model involves static analysis of binaries and only allowing a set of approved APIs.

Apple's static analysis to detect private API usage is not very clever; you can get around it with some fairly trivial tricks.

No, it's not a security mechanism period. Public/Private APIs aren't part of the security model, it's a way to prevent app breakage in their ecosystem due to applications using unstable APIs.

> This is at least plausible to me: to have effective sandboxing, you need to restrict the APIs that a device can call, which means blocking the ability to generate and run arbitrary machine code on-device.

This isn't true. You can sandbox arbitrary machine code by restricting the process it runs in (this is how seccomp works on Linux).

But iSH isn’t running arbitrary machine code, it’s a x86 emulator running Linux inside. It’s more sandboxed than a JS runtime, which actually does compile down to machine code.

It really isn't. You can definitely download code into Scriptable and Pythonista, although I don't think they have package managers.

You've just said it. The difference is the package management. What Apple is defending here is the notion of a store around apps/containers. Apple does not want to provide leaks/wedges into other app ecosystems through native apps. Safari is the only way.

Not out-of-the-box, but users have created scripts that bootstrap package managers into those apps similarly to how iSH users have.