The first
[consensus](https://en.wikipedia.org/wiki/Consensus_(computer_science))
algorithm with
[heterogeneous failures](http://plan9.bell-labs.co/who/garay/continuum.ps),
heterogeneous participants, and heterogeneous observers.
A work-in-progress talk about our Heterogeneous Consensus algorithm.
We use the
[Decentralized Label Model](http://www.cs.cornell.edu/andru/papers/iflow-tosem.pdf)
to show how distributed algorithms, like
[Bosco](https://www.cs.cornell.edu/projects/Quicksilver/public_pdfs/52180438.pdf)
and
[Nysiad](https://www.usenix.org/legacy/events/nsdi08/tech/full_papers/ho/ho.pdf),
can be generalized from more complex trust environments.
We've been looking at modeling distributed system failures with
information flow tools, and expressing heterogeneous trust.
We investigate some potential "slowdown" cases of the classic
[multi-Paxos](http://www.cs.cornell.edu/courses/cs7412/2011sp/paxos.pdf)
algorithm, and developed some new variants to avoid them.