Skimmed the article, grepped for 'hash', didn't find it.
> they are stored in AWS S3
Not sure why anyone would feel the need to couple this sort of thing to S3. Just store file hashes in the database, and have a blobstore mapping hashes to content. I like bitprint URNs for this purpose[1] but any strong cryptographic hash function will do. Then you can use any old datastore you want, be it a plain old filesystem, S3, a Git repo, or something else. It's trivial to bolt on alternate backends because you can always guarantee that what you asked for is what you got. If not, try the next one.
I agree that this is the "better way to do it" but as someone who's done all that, it's a much better DX for whatever system to just take care of this for me. I do think it'd be nice to decouple S3 (hopefully they're S3 compatible, meaning you can switch out minio or R2 or whatever in the back)
> they are stored in AWS S3
Not sure why anyone would feel the need to couple this sort of thing to S3. Just store file hashes in the database, and have a blobstore mapping hashes to content. I like bitprint URNs for this purpose[1] but any strong cryptographic hash function will do. Then you can use any old datastore you want, be it a plain old filesystem, S3, a Git repo, or something else. It's trivial to bolt on alternate backends because you can always guarantee that what you asked for is what you got. If not, try the next one.
[1] http://www.nuke24.net/docs/2015/HashURNs.html