But Haskell has a much better chance of becoming main stream. The type classes allow things like drop-in, easy-to-use hash maps and a simple, type-safe print statement (SML's requires explicit string conversions). Furthermore, GHC consistently pushes modern features; my current favorites are the extremely performant userspace threads (managed by epoll). From what I can tell, the SML folks are more focused on PL research projects than trying to implement a standard library, which is completely fine.
If everything goes well and we as a community build intuitions for strong, static typing, one of these good languages (IMO probably Haskell) could end up as the next Python.
BTW, if you're looking for resources on learning SML, I'd look at Bob Harper's books (http://www.cs.cmu.edu/~rwh/)