Unspecced redacted_by field in unsigned

[joepie91]

Ran into the following event today:

{
	"content": {
		"membership": "join"
	},
	"origin_server_ts": 1595256121167,
	"sender": "@redacted_user_1:example.com",
	"state_key": "@redacted_user_1:example.com",
	"type": "m.room.member",
	"unsigned": {
		"redacted_by": "$redacted_id_1",
		"redacted_because": {
			"content": {},
			"origin_server_ts": 1595261803914,
			"redacts": "$redacted_id_2",
			"sender": "@redacted_user_2:example.com",
			"type": "m.room.redaction",
			"unsigned": {
				"age": 23764322
			},
			"event_id": "$redacted_id_1"
		}
	},
	"event_id": "$redacted_id_2"
}

This event contains a redacted_by property in the unsigned data, but the client-server spec makes no mention of such a field, it only mentions redacted_because.

[richvdh]

as a general rule, the presence of a field in a response from synapse does not necessarily mean it should be added to the spec.

I'm more inclined to consider this a bug in synapse, though not a high priority one.

[joepie91]

Hm. I was interpreting this to be an intended (spec) feature, to reduce bandwidth usage by sending only the event ID rather than the whole event (as is the case for redacted_because) - hence filing it here first. But perhaps that wasn't the intention :)

[richvdh]

(moved this issue from matrix-doc to synapse)

[clokep]

That says:

Servers should include a copy of the m.room.redaction event under unsigned as redacted_because when serving the redacted event to clients.

Note the redacted_because instead of redacted_by, so I think there's some sort of inconsistency still?