Jonathan de Jong
Jonathan de Jong
thought: Option 3 could enforce invariants of retryable and non-retryable responses
> I believe that means we need to allow an m.relates_to field on all room events now. Please get some clarity on that from the SCT.
What does the spec say? Shouldn't it try to send all events it can queue up (within reason)?
Maybe rooms need something similar to `ResponseCache.wrap()`, in the sense that getting the backlog and state_id is handled by one worker, and every subsequent request is queued to that until...
Switching crates like `ruma-client-api` to `no_std` would be useful for implementations such as [ruma/lb](https://github.com/ruma/lb), which could then work in IoT environments.
I think it works better if we have something like `optional_select_path` which contains that bit (before calling to `select_path`), where at macro time we swap out the actual function used....
It would help make the issues more specific to matrix-rust-sdk, as currently the templates from matrix-org generic cannot address things like matrix-rust-sdk versions in a potential better bug template ("please...
Yeah, that issue template was grabbed from the matrix-org organisation .github repo, so adding issue templates specific to this repo would help.
(Hopefully by the time 1.8 rolls around, and we havent seen any calls for something like, we can close this)
While working on this, I realised it was a bit weird to not include the `*_path` variables into this as well, considering that they fall almost under the same invariant....