douglaz
douglaz
I don't think you can disable the `readonly` now. Perhaps better just remove the flag and let it be true?
> @douglaz any idea why you left that comment there? Not sure. I guess there is a risk of getting a inconsistent state from the DB?
> I don't know if its worth spending a ton of time debating this, but there is a tradeoff here between code readability and formatting overhead. Indeed, formatting overhead sucks,...
I thought this would never happen. Congratulations!
> @douglaz 🤷 it is reponsiblity of client integrator, is this issue for fedimint cli? fedimint-cli is one example. If we can't rely on the client to implement it, then...
> Does this make sense for JSON over websockets? In theory, yes, perfect sense because a websocket uses HTTP upgrades. In practice it's tricky because browsers limits http headers for...
> Unfortunately I haven't found a way to support that with `jsonrpsee` yet, so maybe we'd need to upstream some changes first. It should work using tower. I'll test it...
@elsirion It's not fully working, but this should be the basic idea: https://github.com/fedimint/fedimint/pull/2455
But the issue is that now we'll require the whole api to be authenticated. For instance the whole `ConsensusApi`. But, as far as i can tell, the peers don't have...
> > But the issue is that now we'll require the whole api to be authenticated. > > I think that could be solved by only having a subset of...