Sadly, I haven't collaborated yet.. but I would use svn for collaboration. If it's supposed to be a top-secret collaboration, then I'd go for a commercial web svn repository (lots out there that are very cheap). If you think no one could really steal the work from you or that it's ok for people to read it, I would use a free svn repository. The pain is when your partner isn't a techy person, you can't even ask him to merge new versions of works with some sort of merger.. you are supposed to do all the merging from his end.
As for myself, I would not use svn. I usually have papers 15 pages long or so, and I just keep backups everytime I make a milestone and then zip the whole thing (like every month or so).. in case something goes wrong I go back to the backup, it's like a small scale inefficient svn.. but you don't want to kill a bird with an atomic bomb. Always worked for me, I'm not sure for writing books.. probably an svn would become necessary, but I won't be suprised if my backup technique work as well.. it worked for my PhD dissertation anyway :p
EDIT: Most of us mathematician wouldn't have our entire .tex work exceeding 200Mb. For that I would recommend the svn web repositories like (these support closed source, i.e. private projects): XP-Dev and Unfuddle These two are free if you don't exceed 200Mb and you can choose to have the project (you are allowed max 2 or 1 projects for these free repositories.. but I would just keep them in one big folder and name them as 1 project). I like the look and feel of Unfuddle, but XP-Dev isn't that bad either. The advantage of SVN is that it caters for both windows and linux users (for windows I'd recommend TortoiseSVN Client). In case you exceed the 200Mb (which for our purpose is hardly believable.. unless you have lots of images in your LaTeX documents), then I would recommend Assembla because it's very cheap (like \$3/Month/user + \$0.3/100Mb/User) and reliable and you can buy as much space as you want.