One of the more interesting ones was from John Cleave of Experience Builders which is probably worth the aggrevation of having to register, but he definitely points to some interesting ways in which composition can help that line up with what I've been talking about. He also comments:
Move away from the packaging model. If my course consists of an amalgamation of 6 different feeds, and I want to continuously update and improve it as per Flickr, does it make sense for me to have to package and upload the thing to the LMS server? How about I host it, the LMS routes to my site when a user wants to take the course, and my course will simply send the user tracking info back to the LMS as needed.
Push away from “objects” and toward “layers.” The SCO model envisions a world where there are self-contained e-learning units that can be freely mixed and matched. Yet, these are done in a chain: users work through SCO1, then SCO2, and so on. If you look at some of Reilly’s examples and those above, we might have to start thinking in layers: you’re in SCO1 and SCO2 simultaneously, and SCO3 comes in and goes out as the user needs it. The SCO could be a link, an RSS, a data entry form, or even a package of algorithms, never seen by users, that govern communication between sim1 and sim2 and determines when to bring up coaching as contained in SCO3.
Which raises the same kinds of questions being looked at by the LMOS folks.