There are some problem need to be resolved with this solution

• Each page might have different initial state, which ideally, we need have different initial-state.js file for each entry rendered
  • In this case, it will generate too many additional files which will lead to addition network request, and most of the time, they are just setting empty object.
• Currently assets/initial-state.js is not hashed - it will lead to incorrect cache state on site updating.

So, I would think this external them as scripts might not be a very good solution. A few ideas:

• We use nonce to ensure inline script are safe with CSP
• We smartly detect if the initial state is an empty object, if so, we emit the inline script generation - so most of the case, if you don't actually use this feature, it's safe for you.

@antfu Sorry it took so long to answer. I had a lot on my plate.

I agree I have not though about these complications. It definitely makes sense to not use the initial state if it is empty. However, wouldn't it be fine to issue a network request for an initial-state.${hash}.js file for each page if an initial state exists? Nuxt.js does it the same way.
I have not worked with nonces on script tags before, but you would need a CSP Response Header with a different nonce on the inline script on each HTTP request, and I am not sure how you would do this with a statically generated site.

@antfu The initialState is wrong, we should use the initialState inside the promise map instead global state for SSR. This should fix the initialState problem: always use the initialState on first route invocation (on SSR) outside the promise map.

On line 91 we should remove initialState and add it on line 121 (on this PR):

// line 91
const { routes/*, initialState*/ } = await createApp(false) // <= `initialState` should be removed

// line 121
const { app, router, head, initialState } = await createApp(false, route) // <= `initialState` should be added

I don't know if you're fixing it and I don't want to bother you, I added this comment a few days ago: #72 (comment)

@pantajoe Can you test with those little changes on local?

@userquin Yeah, I'll test it and let you know if it solves the issue in the comment you linked.

@antfu @userquin Due to the discussion in #72, let's put this on hold and implement this together with the outcome of #72.

@antfu Sorry for the spam, but on second thought, I believe we can utilize this feature as the initialState now is compatible with CSP. And due to the discussion in #72, it's best if the initialState is only provided globally and not per route.
The isolated page state is another feature that has to be implemented differently and separately.

@userquin What do you think about this?

@pantajoe I think it is a breaking change, the problem is that the changes on #72 are to fix the expected behavior, it is a bug on vite-ssg.

I suggest you wait the input from @antfu ...

I have raised a separate pr #108 to resolve #72. It solves my issues and will help here as well.

Thanks @userquin for the proposed fix

@pantajoe I think it is a breaking change, the problem is that the changes on #72 are to fix the expected behavior, it is a bug on vite-ssg.

I suggest you wait the input from @antfu ...