Benjamin Saunders

> Perhaps we could change the API so that [...] This seems like excessive complexity to me, considering the small population of QUIC implementers, and the assorted other verifications that...

I feel like dynamically checking the correctness of `MaybeUninit::assume_init()` is a large part of what rust code wants from msan.

This would be awesome for QUIC!

> Server certificate verifier happens in the middle of the handshake. I think the same principle stands: create an object for an outgoing connection, drive I/O on it until it...

It's worth noting that separate types aren't *necessary*, they're just a lot less error prone due to correctness by construction.

Isn't the proper way to logically extend a session to issue a brand new ticket on the resumed session, rather than permitting reuse of the ticket any number of times?

> though they don't seem to implement the "soft" limit idea, which seems wrong? Can you elaborate on this? Quinn triggers a key update immediately when the number of packets...

> One difference is that it requires the QUIC implementation to do the actual record encryption/decryption, whereas Rustls encapsulates the encryption/decryption. This is a bit of an oversimplification. Rustls implements...

It's worth note that "is human readable" isn't *quite* the same thing as "is a format which can skip fields", but it's reasonably close. An alternative solution would be to...

> serialize a default value if possible I think serializing the actual value when skipping is unsupported would be much less surprising (and avoid imposing a `Default` constraint).