Mixing facts and dimensions

Here's a simple question: What are good arguments against mixing fact and dimension tables in a star model? or more specifically, against adding dimensional attributes to a fact table to avoid having an extra dimension with very few attributes? Thanks!

1. It would no longer be a star model.

2. Performance would suffer by turning a relatively small dimension table into a huge fact table.

Thanks for your reply. I can imagine that we might also run into problems when trying to extend the model and add more facts or attributes.

Think type 2 dimension. Can't really do the same functionality with all data in one table. People always want to see how the data was when the event happened and how it would look with today's dimension values.