I especially agree with item #2. I have been involved with several web based projects that have "failed" only because someone loosely associated with the project put a release date and the dev team made the mistake of agreeing (it seemed like a sure thing at the time). However, this deadline took its toll on the team's moral. As it became more and more evident that we would not meet the deadline we became more and more grumpy with each other and started shooting down great product ideas because the couldn't be done NOW. It resulted in lots of bad blood and several resignations.

That being said - I do think there are certain situations where release dates will exist in a SaaS product. For instance, if you market tends to start and stop activities on a set calendar (ie: school year) or if there are a handful of trade shows that are critical to demonstrate new features at. I think it is about being insanely pessimistic about how much you can actually achieve. Figure out the minimum you need to demo at a date and work on that first, but never commit to a set of UI diagrams with features that are not insanely flashy for a demo.

my $0.02