I'm with Dion 100% on meeting the needs of users, but as I said on the list, I consider the benefits of using 1.4 – allowing us to use the latest javamail/activation versions, much better charset support, etc. – outweigh the new dependencies. It's better than asking users to stick with the known bugs of javamail 1.3.1 and being forced to reimplement the charset code in 1.3 when the majority of people using it are on at least 1.4.
It's not my project, though, and if 1.4 is unacceptable for some other reason, we'll think of something clever. As Piero says, the important thing is to get fixes out for what's not working.