This sounds like it's going to start becoming a major inconvenience to
you before long. I admit that I haven't had any more thoughts about a
good way to deal with this issue before Pango is ready, have you?

This gtkWindow idea came to me yesterday after reading some code sent by Peter Clifton to the cairo-devel list and I consider it the best approach I could recently figure out. It is not that inconvenient. Indeed this is a lot useful since now I can see some rendering, and I can focus more on the task itself instead of boilerplate constraints.

If I work nice on the cairo world, I guess there wont be much problem when incorporating a future pango support for cairo-userfonts, do you agree?