Result of mutually-distrusting Ids depends on order they're processed

[kornelski]

If user A distrusts user B, and user B distrusts user A, it's indeterminate what result this will have on the WoT:

• if A is processed first, then B is excluded, and A isn't distrusted.
• if B is processed first, then A is excluded, and B isn't distrusted.

[dpc]

The graph flooding algo that is currently used is processing things according to trust level and distance. So whoever is more trusted will cancel less trusted proofs first.

There's even a test, IIRC

cargo-crev/crev-lib/src/tests.rs

Line 282 in 291abd6

fn proofdb_distrust() -> Result<()> {

[kornelski]

What if both are with trust = high?

[dpc]

The exact order is determined by

cargo-crev/crev-wot/src/lib.rs

Line 1052 in a1ba22d

struct Visit {

. The last one is Id which at least makes it deterministic.

[dpc]

The graph flooding algo that is currently used is processing things according to trust level and distance. So whoever is more trusted will cancel less trusted proofs first.

Actually this is not the case. Two IDs at the same effective trust level, distrusting each other will remove both from the trust set.

cargo-crev/crev-lib/src/tests.rs

Line 324 in 291abd6

// Both should be removed from the trust_set.

Kind of a Mutually Assured Destruction.

[dpc]

Closing, as I still think that's how I want it to be, and some improvements around handling it will be done for #390