Judging just by the title, this should be a very interesting talk!
The migration project (from a legacy COBOL system to a Java SOA system) was tried three times before and failed and it was decided that a big bang migration was not possible. There were several different user groups: office users, outdoor users and other systems. The airport is a 20 million passengers per year airport and there were some time constraints, since the two employees maintaining the old system were scheduled to retire within 24 months. The project lasted 8 months, with 31 employees working on it.
The approach the authors took is to implement a new system ans start a new database and keep them online, next to the old system and systematically migrating more functionality to the new system. Initially, the new system listened on the legacy interface for flight data updates and transforms that to a new model before forwarding it to the primary systems. Then the components were are now merged together. The legacy system is updated asynchronously to keep in sync. Finally, the old system was disconnected all together.
And, it worked! Thanks to (according to the authors) to high user integration and detailed requirements work.