I think the blocks are one part of it, and the other part is that S3 doesn’t support renaming or moving objects, and doesn’t have directories (just prefixes). W

I think the blocks are one part of it, and the other part is that S3 doesn’t support renaming or moving objects, and doesn’t have directories (just prefixes). Whenever I’ve seen something with filesystem-like semantics on top of S3, it’s done by using S3 as a storage layer, and building some other kind of view of the storage on top using a separate index. For example, maybe you have a database mapping file paths to S3 objects. This gives you a separate metadata layer, with S3 as the storage layer for large blocks of data.