[Standards] Message Retractions

JC Brand lists at opkode.com
Tue Sep 24 13:02:44 UTC 2019

On Tue, Sep 24, 2019 at 10:40:30AM +0200, Maxime Buquet wrote:
> On 2019/09/23, JC Brand wrote:
> > Based on our off-list discussion, I'm going to go with sending an IQ to the MUC
> > JID in order to ask for a message to be retracted.
> > 
> > The MUC then sends out the retraction message to all participants. This solves
> > the problem of temporary moderators retracting messages and clients then later
> > being unable to verify that the operation was done by someone with the
> > necessary permissions.
> > 
> > The MUC can then decide to replace the message with a tombstone in its XEP-0045
> > message history or in the MAM store.
> > 
> > I guess for MAM tombstoning the MAM service needs to advertise it via disco and
> > the client needs to specifically ask for it in the IQ?
> > 
> > When it comes to one-on-one conversations, I'm not sure what the equivalent
> > interaction would be given that there isn't a dedicated service as in the case
> > of MUC and because two MAM archives are involved. Perhaps sending the
> > retraction message yourself (instead of asking for it to be sent via an IQ) is
> > the preferred way there.
> > 
> > I'd be happy to hear some suggestions. I'm leaning towards restricting
> > the scope of this XEP to only MUCs (and MIX?) and ignoring the one-on-one
> > usecase entirely.
> What about looking at it the other way around? 1:1 being the general
> case and MUC something that modifies/enhances it.
> In 1:1, one would be able to send a retraction, referring to a previous
> message that has been sent with the same barejid. I think this should
> also be possible in MUCs.
> Additionally in MUCs, one could send an IQ to the room so that the MUC
> itself broadcasts it to all participants of that room.

I don't like the idea of there being two ways of doing the same thing in a MUC.
IMO there should be only one way of doing this, and given that sending the
retraction message yourself can cause problems with verifying validity, I think
it should only be via an IQ.

For 1:1, sending a message yourself appears to be fine.

It's a bit unfortunate that we then don't use the same semantics (i.e. stanzas)
for both use-cases, but I don't see currently how we can reconcile them.

More information about the Standards mailing list