Why is cid null, for deletion events?

[DavidBuchanan314]

Delete commit operations appear in the firehose with the repoOp cid field set to null.

atproto/packages/pds/src/sequencer/events.ts

Lines 62 to 63 in 7cb8c62

	if (w.action === WriteOpAction.Delete) {
	cid = null

IMHO, it would be useful to have it reference the cid of the record that's being deleted.

For "update" operations, it would be nice if there was a way to have them reference both the previous cid, and the new cid.

This would make it easier for clients that are not syncing/storing full repo state (such as feed generators) to verify what's being deleted, and to make sure they're receiving/applying operations in the correct order. (for similar reasons reason that a client might want to pass swapRecord when calling deleteRecord)

[bnewbold]

Here are some middling arguments against including the cid:

• extra bytes on the wire, waste of resources in many cases
• draws less attention to the content which has been removed
• implementation performance flexibility

I don't think we have a really strong reason why we don't include it. I can see that it would be helpful in some cases, though I suspect that the majority of downstream services which are trusting the firehose stream (as opposed to keeping a local mirror of the repo to validate events against) wouldn't even be indexing with the CID, just the rkey.