Small problems with spread: it seems like it's more of a PVM/MPI replacement then universal messaging platform, and should run in perfectly-trusted environment.

It seems like there is absolutely no authentication, and any user can pose as any other user/service, this may be fine in tight cluster, but what would you do if you wanted to send message 'Processing finished' to system programmers or support stuff?

You'd need another messaging platform for that.

Seems like secure spread fixes some of those consernes though...

I wonder why it's so hard to get to protocol specs, all people publish these days are 'APIs'.