Kevin Gibbons

Oh, I should mention also that doing `Object.freeze(Object.prototype)`, while it is coherent from the language's perspective, will also break a lot of pages. This is because of a (IMO) very...

> if it turns out that allObjects (i.e. `Object.freeze(Object.prototype)`) won't work in real applications I promise you that it will not work in real applications; there's no need to run...

> Array has `length` property which isn't function, so that will be vulnerable to PP even with `preventExtensions` `length` is an own property of every array, not a property on...

> Not sure if I understand this. Polyfill will just be `Object.preventExtensions(Array.prototype)`. I mean, people will want to polyfill new JS language features, not this feature. That is, when JS...

@guybedford That's intentional. Otherwise the README's example of [denying access to a built-in module](https://github.com/WICG/import-maps/#denying-access-to-a-built-in-module) would be impossible; see #137 for the motivations behind this PR.

The simple mental model we explained makes sense only when all of the mappings are similarly specific. So, if all mappings are in the same scope or all at the...

> So I'd like to have something -- simpler semantics of cascading, or something else -- that helps the developers of import maps to be confident that their import maps...

@domenic > Add an example or two or three of concatenating import maps, and explain the mental model. This replaces the previous, now-inaccurate example. I would appreciate if @michaelficarra or...

> Er, can this PR already result in exponential-sized output? Hm. As implemented, yes, because there's nothing forbidding having the same specifier appear multiple times in a fallback list (even...

Ah, yes, I keep forgetting about package-prefix mappings. In this case that seems to be what the user is asking for, though, so I'm not sure to what extent it...