Jacob Hoffman-Andrews

Thanks for trying curl with rustls, and thanks for the report! Error 7010 from rustls-ffi is "UnexpectedEof" (TODO: get better output for this error message). Rustls will emit UnexpectedEof if...

I've opened https://github.com/curl/curl/pull/10463 to improve the error messages somewhat, and confirmed that Google's HTTPS servers exhibit the same behavior we see here for IMAPS: the TCP connection is reliably closed...

I agree a user-facing error doesn't solve anything. I was thinking an internal error that would be emitted by TLS backends and handled by the higher protocol layer - either...

Reopening discussion to consider https://github.com/rustls/pki-types/pull/24. Also I wanted to note (and this is clearly a rustdoc bug): the rustls re-exports will show `Available on crate feature alloc only` for some...

> callers would prefer not to link all key exchange groups/cipher suites, so there must be a way to build a config that doesn't call any functions relying on the...

Thanks! And yeah, that makes sense. But say someone wants to not depend on ring or aws-lc-rs at all, which traits would they have to implement? While you're working on...

My initial motivation was that @cpu had mentioned on rustls-ffi wanting to export the CryptoProvider API in FFI, so I wanted to understand the API upstream. For audiences, I've drawn...

> I don't think we should provide FFI APIs that allow custom CryptoProviders to be written. Yep, that's the conclusion I came to as well, in https://github.com/rustls/rustls-ffi/issues/366.

> Sounds like you may have missed https://github.com/rustls/rustls/issues/1372#issuecomment-1816537477. 👍 Aha, thanks for the link. Yes I saw that comment and then immediately forgot the specific recommendation of using CryptoProvider as...

It seems like `rustls_pki_types::SignatureVerificationAlgorithm` is part of the extension surface exposed by CryptoProvider, by way of rustls::[`WebPkiSupportedAlgorithms`](https://rustdoc.crud.net/jsha/rustls-doc/rustls/struct.WebPkiSupportedAlgorithms.html#), right? Do we have a semver hazard here because that trait is in...