undici parses FormData bodies incorrectly

[KhafraDev]

The parsing should be synchronous. The spec makes it abundantly clear that you buffer the body in memory and then parse it, which we do for every other body mixin. It also doesn't make much sense for us to parse the body asynchronously only for us to buffer it into memory later on. It's not meant for server environments and we shouldn't give people a false sense of security.

• It's also extremely slow
• We also fail a large amount of WPTs regarding formdata parsing because of this.
• There are a large number of hacks/workarounds for shortcomings of our current approach.
• Every other platform (Deno, Webkit/bun, Chrome, and Firefox) parse it synchronously. I've previously posted proof of this in an issue but it'd be hard to track it down.

If someone wants to stream the formdata parsing:

const response = await fetch('https://formdata.idk.com')
const busboy = new Busboy({ headers: { 'content-type': response.headers.get('content-type') } })

Readable.from(response.body).pipe(busboy)

[KhafraDev]

https://andreubotella.github.io/multipart-form-data/

[climba03003]

I am supporting it follows by spec.

Only concern is that if it buffered the whole content to memory, what will happen if it exhausted the memory?
An error, or crash? I hope it should be the former.

Can undici provide a handy tools for async consume?

[KhafraDev]

The same thing that happens currently, it'll crash. The body is parsed asynchronously but then buffered in memory right now.

I don't think undici needs to provide a way of efficiently parsing multipart/form-data bodies, busboy and alternatives do that already.

[KhafraDev]

Every other platform (Deno, Webkit/bun, Chrome, and Firefox) parse it synchronously. I've previously posted proof of this in an issue but it'd be hard to track it down.

Found it..

Firefox - https://github.com/mozilla/gecko-dev/blob/7f3ff3f4d34e7d234da4f5f5b345d2add7f30e95/dom/base/BodyUtil.cpp#L67
Chromium - https://source.chromium.org/chromium/chromium/src/+/main:third_party/blink/renderer/core/fetch/multipart_parser.cc;l=1;bpv=1;bpt=0
Deno - https://github.com/denoland/deno/blob/0cd05d737729b4cfab1d5e22077b3b9ad4ed5e30/ext/fetch/21_formdata.js#L490
Webkit - https://github.com/WebKit/WebKit/blob/f0350d6575c366d884d98f9937e77fe499b93398/Source/WebCore/Modules/fetch/FetchBodyConsumer.cpp#L129