Implement DebugSecret for vectors of debuggable types

[ionut-arm]

Hi,

Currently DebugSecret is implemented only for Vec<S>, where S must have DebugSecret implemented on it as well (link) - however only arrays of types with Debug have that trait, so only Vec<[T]> will get DebugSecret implemented.

I was wondering if it's possible to have DebugSecret implemented on Vec<S: Debug> as well/instead.

[tony-iqlusion]

Thanks for reporting this.

For context: one of the goals of this crate (which could definitely be better documented) is preventing accidental secret leakage through mechanisms such as Debug logging.

To enforce that concern, DebugSecret::debug_secret presently returns a &static str.

I think you're right that having DebugSecret contingent on Vec<S> where S: DebugSecret is not ideal. I think instead of asking the downstream type to opt into DebugSecret, instead the type signature for DebugSecret::debug_secret could be changed to use const fn, which would uphold the same invariant of no data-dependent debug logging, but in a way that's more flexible than a &'static str.

By leveraging const fn, a DebugSecret impl on Vec<S> could incorporate the core::any::type_name of S.

Is that what you'd ultimately like for debugging information? If so, it seems possible.

[ionut-arm]

That definitely sounds like a sensible solution that would solve our problem.

Is that what you'd ultimately like for debugging information? If so, it seems possible.

Actually, it's not really debugging that I'm interested in, but serialization. However, Serialize and Deserialize are tied to DebugSecret, and thus my request :) I should've mentioned that previously.

We're trying to protect a secret while working with it in memory, but to also be able to serialize it at the right point. Don't know if that would be better achieved through other means, like exposing the contents of the secret and then serializing, but it feels more error-prone.

[tony-iqlusion]

You might take a look at the docs for SerializableSecret:

https://docs.rs/secrecy/0.6.0/secrecy/trait.SerializableSecret.html

...which, upon further inspection, does not appear to be properly used in the bounds for the Serialize impl on Secret (!)

To prevent accidental exfiltration of secrets, SerializableSecret was supposed to act as a marker trait, and one deliberately not impl'd for SecretVec.

The alternatives are:

1. Make your own newtype that wraps Vec<_> (of whatever type you're interested in)
2. Use serde's serialize_with as noted in the SerializableSecret docs.

All that said, it seems this crate could use a little work and documentation improvements. Hopefully I'll have time to make a pass over it in the next few days to clear some of that up and fix the issues brought up here.

[ionut-arm]

Thanks for pointing that out! For some reason I was completely blind to the existence of SerializableSecret in the crate...

I'll probably go down the route of using serialize_with, seems more idiomatic.

[tony-iqlusion]

#463 should address the issues with the trait bounds: serde serialization is no longer bounded by DebugSecret.