But to start a project without any notion of a deadline is a sure fire way to never release anything. I believe that projects like this require a "happy medium".
Come up with a list of features to implement, estimate those features (and those features only), design, implement, test, release with re-assessments (and iterations) happening throughout the entire life-cycle of the project, adjusting estimates as necessary.
At some point the companies need to publish a release date to the public. That should be somewhere near the end of the project, assuming the re-assessments warrant it.
Do NOT allow new features. Period.