+1, if we're happy with just SHA-1s.

Oh, except:

Isn't it a little strange to put content on disk, and then go back to the disk and ask what the digest of that content is?

Only if you trust your disks implicitly. :)

Hmm. I figured if the request doesn't include a checksum, we can't guarantee that when we calculate the checksum the resource hasn't already been fouled. We may as well worry about flipped bits in RAM

Here's the other down-side.. if the content was quickly spooled off to disk (or, worse, a different node of the cluster), asking for ISPN for a checksum might be unnecessarily slow (I'm not sure how they calculate the hex digest though..)

So this is a checksum on transmission, and the asynchronous fixity service is responsible for checksumming persistence?

Sounds ok to use:

    /**
     * Get the hexadecimal form of the SHA-1 hash of the contents. This hash can be used to determine whether two Binary instances
     * contain the same content.
     * <p>
     * Repeatedly calling this method should generally be efficient, as it most implementations will compute the hash only once.
     * </p>
     *