Currently clients need to guess at which room versions the server supports, if any. This is particularly difficult to do as it generally requires knowing the state of the ecosystem and what versions are available and how keen users are to upgrade their servers and clients. The impossible judgement call for when to encourage people to upgrade shouldn't be impossible, or even a judgement call.
Building off of MSC1753 (capabilities API) and
the recommendations laid out for room versions,
this proposal suggests a m.room_versions capability be introduced like the following:
{
"capabilities": {
"m.room_versions": {
"default": "1",
"available": {
"1": "stable",
"2": "stable",
"state-res-v2-test": "unstable",
"event-ids-as-hashes": "unstable",
"3": "future-scifi-label"
}
}
}
}Clients are encouraged to make use of this capability to determine if the server supports a given
version, and at what level of stability. Anything not flagged explicitly as stable should be treated
as unstable (ie: future-scifi-label is the same as unstable).
The default version is the version that the server is using to create new rooms with. Clients can make the assumption that the default version is a stable version.
Clients should encourage people with sufficient permissions to perform an upgrade to upgrade their
rooms to the default room version when the room is using an unstable version.
Changes aren't pushed to the client, which means clients may want to poll this endpoint on some heuristic instead. For example, clients may want to poll the endpoint weekly or when the user relaunches the client. Clients may also wish to provide users a way to upgrade without considering the capabilities of the server, expecting that the server may not support the user-provided version - the intention being such a feature would be used by eager room administrators which do not want to relaunch their client, for example.