Example: Dining cryptographers #102
Conversation
This comment was marked as off-topic.
This comment was marked as off-topic.
| R2Err, | ||
| } | ||
| // TODO: for protocols that do not have "provable errors" we should have a shortcut and/or blanket impls. | ||
| // TODO: Why does ProtocolError require Display? |
There was a problem hiding this comment.
Basically, to show some short version of the error in a debugging/logging scenario, where an actual Debug would be too long (because it may contain auxiliary information)
There was a problem hiding this comment.
I'd argue that it's prooobably not our business (as library authors) to decide if a Debug output is "too long" and imposing Display as a stand-in for Debug is a bit odd.
| /// bla | ||
| R2Err, | ||
| } | ||
| // TODO: for protocols that do not have "provable errors" we should have a shortcut and/or blanket impls. |
There was a problem hiding this comment.
There is an empty NoProtocolErrors that implements ProtocolError.
|
|
||
| type ProtocolError = DiningError; | ||
|
|
||
| // TODO: Having blanket impls for these methods would cut down on the boilerplate a bit. |
There was a problem hiding this comment.
True, but what would be the default behavior? If it's just ignoring the error, there is a danger of forgetting to implement them.
That said, I am not happy with these methods. They are far away from the message type, and not very convenient to write and debug. There should be a better way - somehow bundle these checks with the message type itself.
There was a problem hiding this comment.
You mean: protocols are so different that there's no reasonable default here? That's fair I guess. OTOH there's a limit to the amnount of hand-holding a library can provide beyond "make it easy to do the right thing". Tricky to know, but I guess the it's an indication that writing examples is useful to see where the sore spots are.
| payloads: BTreeMap<Verifier, Payload>, | ||
| artifacts: BTreeMap<Verifier, Artifact>, | ||
| ) -> Result<FinalizeOutcome<Verifier, Self::Protocol>, LocalError> { | ||
| // Prolly just a NOP in this protocol? Maybe its artifact are the two bits (one private, one shared)? |
There was a problem hiding this comment.
Since you have the second round, it should at least return Ok(FinalizeOutcome::AnotherRound(BoxedRound::new_dynamic(Round2 { .. }))) (yep, ergonomics is not great here. Can be improved)
| } | ||
| } | ||
|
|
||
| // TODO: I don't understand why there needs to be a Signer and a Verifier type. Suspect it's a CGGMP "contamination" and that they are not needed at all for this protocol. |
There was a problem hiding this comment.
It's rather that manul assumes you need authentication and evidence generation for your protocol (which you generally would in a production setting). Not needing those leads to unused arguments.
There was a problem hiding this comment.
You can also use manul::dev::TestSigner/TestVerifier which do pretty much what you've done here.
There was a problem hiding this comment.
Yeah, I tried to do without using manul::dev but ended up doing pretty much exactly the same thing.
DRAFT
Implement Dining cryptographers problem in
manul, as an example of a non-trivial protocol.