spec does not say that invalid m.relates_to properties are rejected

[richvdh]

If someone tries to send an event with an m.relates_to property which is not an object (for example, m.relates_to: null, per matrix-org/synapse#5404), should the server reject the attempt with a 400 error, or should the m.relates_to be ignored?

[turt2live]

In practice it should be whatever the server would do for the wrong kind of body.

[richvdh]

In practice it should be whatever the server would do for the wrong kind of body.

Synapse currently returns a 400 with: {"errcode":"M_UNKNOWN","error":"'body' not a string type"}. That's probably a reasonable precedent to follow here.

[richvdh]

this is touched on in MSC2674, but still not really answered; see matrix-org/matrix-spec-proposals#2674 (comment)

[clokep]

I think I implicitly changed the behavior for this in matrix-org/synapse#11161 where it will just ignore the relation instead of error. 😢

[richvdh]

this is touched on in MSC2674, but still not really answered; see matrix-org/matrix-spec-proposals#2674 (comment)

This got resolved during FCP on MSC2674; it says:

Servers should reject events not meeting these conditions with an HTTP 400 error when they are received via the client-server API.

However, the specced text seems to say something different:

Relationships which don't match the schema, or which break the rules of a relationship, are simply ignored.

So I think this is now a spec-bug.

[turt2live]

fwiw, the "ignore the thing" bit is still accurate to the MSC (it says that clients which receive an invalid relationship are to treat it as though there is no relationship). It's just the 400 error which didn't make it.

[richvdh]

sorry, yes.