In our statement about the role of the Bersekas-Tseng codes we say we use both RELAX-IV and AUCTION. In point of fact, we've always set things up in such a way that AUCTION is never invoked. Perhaps we should leave that out.
Another approach would be to expose from within R the RELAX-IV parameter CRASH that directs RELAX-IV to use AUCTION for the initialization. It would be relatively easy to do this at the level of signature of the internal function fmatch. (I have no idea what would happen if we tried this. I'm somewhat fearful of it uncovering that I somehow broke the option for AUCTION initialization years ago.)
Whatever we do about that, let's add citations to Bertsekas's Linear Network Optimization (1991, MIT Press) and Network Optimization: Continuous and Discrete Models (1998, Athena Scientific).
CRASHthat directs RELAX-IV to use AUCTION for the initialization. It would be relatively easy to do this at the level of signature of the internal functionfmatch. (I have no idea what would happen if we tried this. I'm somewhat fearful of it uncovering that I somehow broke the option for AUCTION initialization years ago.)