Abstract
Distributed constraint satisfaction (distributed CSP) addresses the problem of assigning values to a set of variables, to satisfy the secret constraints of several participants. Many have argued so far about how to formalize the privacy requirements on input constraints. However, we notice that not sufficient attention was given to formalizing the description of what solution is desired. Different criteria of selecting a solution have strong consequences on an inherent privacy loss. We conclude proposing a way to minimize privacy loss by specifying expectations about how a solution has to be chosen among possible candidates.