Thinking about how to move the TCP-AO intent forward: clearly a 36-patch set is a bit much. The first 6 patches are prep work, and we know there is a use case for those. We could handle patches 3 and 4 as a stand alone set first. Once merged, deal with the crypto API and users until those maintainers are good. That would be patches 1, 2, 5 and 6. Once those are merged it drops down to just networking patches with selftests. Those can be split into AO (19) and selftests (11) making it 4 total sets of manageable size. The AO patches can be reviewed until convergence on a good starting point. Sound reasonable?