refs: Prohibit IDs consisting of all zero bytes

[cthulhu-rider]

i propose to reserve zero IDs and prohibit them from being message fields. While this is a breaking change in theory, in practice nothing will break. At the same time, application development will benefit

[carpawell]

Overall agree. But do we really need to specify it for the fields that are commented as "hash"? What if container's hash is all zeros? It a caller is a liar (and that is not a 32 zeros hash), it is easy to check.

[cthulhu-rider]

What if container's hash is all zeros? It a caller is a liar (and that is not a 32 zeros hash), it is easy to check.

we see zeros - we dont hash the container at all. That's the difference

But do we really need to specify it for the fields that are commented as "hash"?

for hash - mb not. But i prefer to explicitly state such cases. For example, although zeros are not a particularly valid hash, the protocol may dictate a special interpretation of them. And so we emphasize that it does not exist

[roman-khimov]

Refs. https://datatracker.ietf.org/doc/html/rfc9562#name-nil-uuid

[carpawell]

@roman-khimov, I have checked UUID generators we used and it is impossible to get zero ID. At the same time, I do not understand what prevents getting zero sha256.

[roman-khimov]

The probability of getting it is rather low. Hash collision between two containers (or objects) is more probable than that and if it happens -- we're toast.