It is claimed that making the protocol and source closed lets the author innovate new features more easily. This needs more careful explanation, especially as having an open protocol doesn't necessarily have to hinder innovation or implementation of new features.
You could just publish RFC for the protocol and let others worry about keeping up with changes. If you want to be supportive you could possibly hold off new changes for a week or two to give them time to upgrade, but even if you don't it's not that big a deal and no real reason to keep the protocol hidden.
After all, even if it *is* undocumented, it's not too difficult to decipher it with the right tools.
As long as you are not charging anything for the client (or have plans to do so in the future) there is no good reason for keeping the protocol hidden or go about blocking people from using 3rd party clients they've written themselves.
If someone wants to cheat, they can do it already, so not even that is a reason.