Once in production, sometimes you want the user's browser to serve these local files to itself, from blobs and/or the local file system, meaning a server or CLI isn't feasible.
This should be better from a privacy standpoint since files stay completely client-side, but CORS difficulties end up encouraging developers to push files to the server at least temporarily instead.
Isn’t this privacy tradeoff directly at odds with security? i.e. local access would also open the door to malicious sites accessing the local file system?
I can understand the dev-time frustration with CORS, but removing it just reintroduces a whole category of security issues.
Maybe there is something better, but whatever replaces it would need to include similar restrictions.
Look at docker for analogy. Sometimes you just need a volume to use locally, but you don't necessarily need to access files that already exist. Consider the use cases of /tmp/ as well, for instance.
But setting that aside, bypassing CORS to achieve this seems analogous to unlocking your front door/gate before leaving for the day in order to grant access to a delivery driver.
It’ll work, and the driver can deliver your package, but it’ll also let random and potentially malicious passers-by into your home without restriction, so who knows if your home will be intact when you return. A theoretically functional solution that doesn’t really work in the real world.
Some other solution that behaves more like a temporary file store sounds better, but the tradeoff I mentioned is specifically about CORS.
I get that you need the user to explicitly select the file but I don’t understand the need to upload to a server.
Couldn’t you ask the user to select the file and just use the file locally, e.g. with the File API? https://developer.mozilla.org/en-US/docs/Web/API/File
I guess things break when you need repeated access and when having to pick the same file multiple times would be bad UX.
In the case I was referring, we loaded files as blobs in the browser's local storage, allowing us to run a web worker as a persistent local cache server.
This involved managing Firefox's admirable response doctoring policies, though.
All other browsers would let a web worker modify the incoming response to a request before forwarding it to an iframe, but FF absolutely refused it and required that the modified response get constructed anew in its local context, meaning locally created or injected content could not be mistakenly trusted as origin content.
`serve ./` turns any directory into a localhost webserver
What OP obviously meant was a way to disable same origin policy checks.
It isn’t anything of the sort. If they think the problem is CORS, they are going to put all their effort into trying to figure out how to disable CORS. This will not help them in the slightest. What they want, in effect, is to have a configuration that is as if CORS is on all the time, which they would never think to do if they think it’s CORS stopping them from doing what they want.
> What OP obviously meant was a way to disable same origin policy checks.
Yes, and they showed absolutely no knowledge of the fact that the SOP even exists. They think it’s CORS doing it. Pointing out this is backwards and pointing out what is actually causing their problem is helpful, not pedantic.
Localhost cannot (unless you set it up that way, which requires user action)
That’s why.
What I want is unrestricted access for my code, and that the sites keep the sandbox.
What I do nowadays is start a server that has basic auth and zero cors, that I can send commands to from my scripts, like, fetch me this resource without cors, or download this to this folder, etc.
So, ex, if site A opens a tab on site B (and so gets a reference to it), site A can read and modify anything on that page. Or, ads loaded in cross-origin items could read and modify anything on the containing page.
It's fine for testing, but don't log into any real accounts in this profile!