Snapshot IDs

[radekg]

Per: https://yugabyte-db.slack.com/archives/CG0KQF0GG/p1644257955047349?thread_ts=1643868640.696389&cid=CG0KQF0GG

Please interpret the resp.snapshot_id() as an array of 16 bytes. And nothing more. (Not md5/base64/etc.) 16 bytes. Each byte value is in: 0x00 - 0xFF range.

In C++ code the decoding only checks the string size and do memcpy:

Uuid Uuid::TryFullyDecode(const Slice& slice) {
	if (slice.size() != boost::uuids::uuid::static_size()) {
		return Uuid::Nil();
	}
	Uuid id;
	memcpy(id.data(), slice.data(), boost::uuids::uuid::static_size());
	return id;
}

Per: https://yugabyte-db.slack.com/archives/CG0KQF0GG/p1644258454859919?thread_ts=1643868640.696389&cid=CG0KQF0GG

Just because the Snapshot ID for non-transaction-aware-snapshot (old snapshot when "transaction_aware=false" - not used now) was passed through the same PB field. So, in the code "32-bytes string" = old non-transactional snapshot UUID as a string. "16 bytes" = new transactional snapshot id in binary form. So.. two-in-one.. it's the reason of the complexity. Sorry.

As the old (non-transactional) snapshots are not used more, you can always expect the 16 bytes.

The case is only for Snapshot ID.

Namespace/Table/Tablet ID is a UUID in simple string form. No such complexities.