You may have accumulated a lot of junk that you don't use and don't want, and it is stored in your legacy repository. In addition, very few projects will find sufficient value in being able to access the historical data from the new repository to offset the costs (both monetary and time) of the migration. Most projects are better served taking a very limited set of revisions (snapshots) into Subversion, but retaining some ability to access the historical data (think archive) in a slimmed-down architecture for that product.