The Challenge of Large File Checksums

ppppp.dev

I don’t understand the goal here.

Splitting a file into chunks, hashing them in parallel, and then hashing the resulting hashes is certainly a valid method but not the same as hashing a file the traditional way.

Unless the world changes how they publish hashes of files available for download, I don’t see the point.

The reasoning here is to improve getting hash of a large file (let say size > 100GB). Reading the file content sequently and hashing it will take a lot longer

I agree, but there is no way to compute the equivalent of the sequential hash using any parallel method.

This isn’t like gzip which can be parallelized.

Without standardization of a parallelized hash computation, it’s just a toy exercise in an embarrassingly parallel problem.

Why not use a faster hashing algorithm like xxhash?

This code is using sha256 which, whilst cryptography secure, is a massive burden for computation.

yup, I agree. I should update it to use a faster hashing algorithm. That is the next step

Reply on news.ycombinator.com