Re: [Monotone-devel] [RFC] versioned policy -- bootstrapping

[Bruce Stephens]

Nathaniel Smith <address@hidden> writes:

[...]

> However, consider an antagonistic fork:
>      A   <-- states that Alice and Mallory have commit access
>      |
>      B   <-- states that only Alice has commit access; signed by
>      |       Alice
>      C   <-- states that only Mallory has commit access; signed by
>              Mallory

Wim's simplification is surely right: edges in the raw tree can be
checked, so the edge from B to C should simply be ignored.  Instead,
Mallory would have to commit a child of A.  And that obviously creates
a branch with two heads, so we can immediately see the conflict.

I've no idea how to fix the more complex problem you describe later.
Maybe SPKI has some ideas?  (Or RFC2704?)

[...]