Mastodawn

#LazyWeb query about why a #YouTube #download isn't deterministic.

OBSERVATION:

Download the same video multiple times using `yt-dlp URL`. The contents aren't the same even minutes apart. The SHA256 hashes don't match.

QUESTION

Why does this happen?

SPECULATION

- yt-dlp doesn't strip out sponsors in a deterministic way

- YouTube tweaks the bitrate on the fly despite requesting the highest quality

- YouTube adds steganographic fingerprints to thwart downloaders via state violence

> QUESTION
> Why does this happen?

First, define your expectation. What, exactly, would you expect YouTube to respond to just one request at that URL?

I'm guessing you know better than to define it as "exactly the video data that was originally uploaded to YouTube".

But then, what data exactly do you expect to be served?

And, second: Do we expect that to be the same data served on multiple requests to the same URL? Why or why not?