Seconded. If you really want a 100-year language, then your chief concern is getting it widely-adopted, and a permissive license like MIT is just the ticket.

I'm not so sure that liberal licensing leads to popularity. One obvious counter-example of your supposition is GNU/Linux vs. BSD.

Incidentally, the last thing the FSF is worried about is popularity, yet look at how popular GCC, Emacs, bash, and other GPL'd software is.

Licensing your software with a very liberal license may help make your software more popular with some large companies who are in a position to take advantage of it (and I suspect timing is important here as well), but I'm not so sure it makes your software popular with the developer community.

It's certainly complicated, and depends on several factors.

One thing that's important for languages like Tcl is that they can be embedded within other projects, and for that reason a liberal license was very important. Of course, it's certainly not the only thing you need. The GPL is probably ok for "finished product" sorts of things that you don't need to link against to utilize it.

I prefer zlib or boost licenses which make a distinction for binaries. You don't have to attach the license for binary-only applications which makes it more suited in some cases (like embedded development).

I second zlib. It's slightly more restrictive than MIT, but in a very sensible way. I don't think preventing modified sources from masquerading as the original Arc, or maintaining author attribution will impede adoption overly much.

Pardon my naivete, but are masquerading languages a common problem?

Choosing a license to fight imagined evils seems like the short route to a DRM'd license.