| 3 Jun 2019 |
 Maximus | See this wiki entry. Please read it fully | 23:10:04 |
 mextor | thx reading now | 23:11:19 |
| mextor | ok, then we can't remove email address ? | 23:25:50 |
| mextor | I guess we can do it manualy on server's side right ? | 23:28:42 |
| Maximus | mextor: the document shows how to do it if you really want to, but 3PID removal is not supported since there is no spec for it | 23:30:37 |
| mextor | I did add :
location /_matrix/identity/api/v1/unbind {
return 404; | 23:32:04 |
| mextor | but I still get the same message | 23:32:19 |
| mextor | btw, my server is not federated | 23:32:33 |
| Maximus | then you'll need to ask the synapse people why it doesn't work | 23:32:47 |
| Maximus | mxisd only support the stable spec, and makes sure that itw remain legal under GDPR. For now, 3PID removal is not in the spec, and we just documented what works for us, but we do not support it | 23:34:17 |
| mextor | ok, thx can I do it manualy in the DB ? | 23:34:52 |
| Maximus | At your own risks. I'm not up to date enough with the synapse internals to tell you yes or no | 23:36:27 |
| mextor | then I only get a user blocked with an unchangeable email ? | 23:37:00 |
| Maximus | It seems like. synapse forces this new behaviour. Previous versions respected the spec and removal was possible. | 23:38:20 |
| Maximus | The wiki document points to several synapse and spec issues you should express your concerns on, including closed ones. We have warned about such breakages and issues already, we were not listened to. Only you can, as a user, raise awareness to the issue. You can also open a new issue on the synapse repo | 23:40:13 |
| mextor | ok, thx Maximus , I will | 23:41:25 |
| mextor | meanwhile is there a command line to completely erase a user, I will recreate it for testing | 23:41:50 |
| mextor | ? | 23:41:53 |
| Maximus | In reply to @mextor:matrix.org
meanwhile is there a command line to completely erase a user, I will recreate it for testing I'm not aware of one | 23:42:19 |
| mextor | thx | 23:43:24 |
| Maximus | even the deactivate API will fail since it also tries to remove emails the same way and will also fail | 23:44:12 |
 TravisR | Maximus: just a head's up that June 10th is Matrix 1.0, which includes an identity server release. That release will include MSC1915, the unbind API we previously talked about. This means that mxisd's behaviour on June 10th will be non-compliant with the spec. Thought I should give you a head's up so you can implement a patch before then. The spec is currently written as https://github.com/matrix-org/matrix-doc/pull/2046 | 23:58:54 |
| 4 Jun 2019 |
| mextor | thx | 00:00:21 |
| Maximus | TravisR: I'm aware. No patch is needed since mxisd doesn't store any 3PIDs. If our feedback was taken into account, spec and implementations will deal with it in an optional matter, just like /3pid/bind does. Up to Riot to prompt the user I guess | 00:04:25 |
| TravisR | The spec has a fallback for identity servers which do not implement the endpoint, or do not want to. | 00:05:42 |
| Maximus | I don't follow how it will help the current problem of synapse failing to remove a 3PID. Or should mextor wait for synapse to implement the new spec release? | 00:07:55 |
| TravisR | Synapse already does implement it. As far as I know, mxisd doesn't return the right combination of HTTP error codes and response bodies to prevent it from screaming. | 00:08:51 |
| mextor | g night guys, its 2am for me | 00:12:37 |
| mextor | back later | 00:12:46 |
| Maximus | TravisR: I'll see what the v1.0 spec has to say about the endpoints (CS and IS) when its out then. I hope it effectively takes into account mxisd use case and doesn't involve GDPR illegalities, and that synapse/Riot behaves properly. mxisd cannot compromise on privacy or promote illegal behaviour that may get people sued for leaking private & identifiable data. | 00:23:31 |
