|
Föderation EN Do 26.06.2025 10:03:14 Dear @matrix , what is the point of maintaining specs of a protocol, if you then decide to break it on matrix.org server so all client apps have to choose between following the specs XOR being compatible with your server? If the Foundation is really the steward of the protocol, then do something about this. It is happening way too often. Currently here https://spec.matrix.org/v1.14/client-server-api/#client-behaviour-9 |
|
Föderation EN Do 26.06.2025 11:14:30 |
|
Föderation EN Do 26.06.2025 11:17:33 @andybalaam @matrix Unfortunately no, this is not a synapse problem, only matrix.org instance problem and I don't see any way to report issues just for that instance. Context: spec says that there is a POST route for deleting devices. Client apps implement it. Now matrix.org decides to do things differently without telling anyone. This route returns 404 from matrix.org, but works on any other synapse. Some people in some random chat rooms discovered that we are supposed to use account.matrix.org |
|
Föderation EN Do 26.06.2025 11:31:35 @andybalaam @matrix Here Cinny chat recently encountering the issue with "tulir" (seemingly representing matrix.org?) confirming that this break of all spec-following clients was done by design Medien: 3 |
|
Föderation EN Do 26.06.2025 11:36:23 |
|
Föderation EN Do 26.06.2025 11:41:48 @chebra @matrix Maybe a bug logged against the spec is the right way to track this? https://github.com/matrix-org/matrix-spec/issues |
|
Föderation EN Do 26.06.2025 12:13:50 @andybalaam @matrix So would you say that this is the bug of specs? So in case of differences client developers should take the matrix.org server as canon? Because I still don't think that's the right way. Matrix.org server is the one failing to implement the specs here. |
|
Föderation EN Do 26.06.2025 12:18:06 @chebra @matrix it looks to me like a bug in the spec change that introduced OIDC for authentication: as part of that change, there should have been a deprecation of this endpoint, because it doesn't make sense in this context. I'm not saying that matrix.org should be the de facto standard, I'm saying that (I think) there is a bug in the real standard, which has become visible when matrix.org implemented it. |
|
Föderation EN Do 26.06.2025 17:56:32 @andybalaam @chebra oops, yup, we're looking into what happened here on the spec side and will get it fixed. |
