I'm completely behind the idea of using Moodle as a guide to the writing of user documentation, too, and I've long announced my intention to work on it.
The thing is, the one-page tutorial that I and my colleagues have used to lead students through a 45-minute hands-on tutorial is sufficient training for logging on, uploading and downloading texts, and contributing to the forums, but the problem of successful course use is much larger. I've conducted this tutorial for hundreds of students, and I find that well over half figure it out quickly and have time left to help the others, and that the others, those with difficulty, have more general and fundamental problems understanding the use of database-drive web applications and even basic PC use.
For these people, I have come to think, the problem is both conceptual and social: often quite brilliant in other fields, things like cookies for them are as incomprehensible as colors are to the colorblind, and no matter of written explanation that I might devise is as nearly a sufficient remedy as their learning how to learn and solve these problems on their own, and on their own means finding someone else they can call and depend on to help them: computer use, at this stage, is as scary as it was for me at my middle age to go to a ballroom dance lesson and step out onto the dance floor(!).
I've discussed this with my Dotlrn colleagues to some length, and while we feel that good documentation is important, the larger issue is developing a supportive learning environment and user community, that we do that with "boots on the ground" before we try to do this online, and that we associate the technological problem immediately with the academic professional one.
We are presently designing a class on "information literacy" whereby we offer a hands-on workshop on the use of the web for academic research and do so as part of developing the library's web-based reference desk services that instructors may then draw on as a resources when designing and managing their classes. The logic here is that the first thing instructors are doing is storing lots of texts online, that this is a traditional reserve desk function, and that librarians also know about catalogueing, and once students have visited the library for their readings they return to visit the reference desk when they need to write their papers. The other logic is, rather than asking instructors who say they have no time to learn something new, we give them a good reason to send their students to us (we have a student/professor ration of 60:1, so instructors are only too happy for the help!)
The thing is, only about a third of our students have pc's at home with fast online connections, most do not feel comfortable with the technology, and that the larger problem maybe best thought of in terms of "literacy" -- which we take to means a certain fluency in a wide range of situations where pc use and communications might be central to academic work.
Everyone talks about how important the web is, but apart from the one or two things people use, this is more hope than practice. We think most of our students and faculty are really at the beginning of understanding how to use this stuff. We no sooner teach them how to use Dotlrn then we run into a wall. Beyond posting syllabi and downloading texts, use of communications and community functions depend on classroom management and intellectual community styles, goals, and development strategies. And how to develop these styles, goals, and strategies using the technology is comparatively undeveloped, and moreover, I think, so long as we see "documentation" in terms of mere functionality we perpetuate the view that the course is here, the techology is over there, and you have only to bring the students to the technology -- and not bring the technology to the students, their courses, and their life world.
Which brings me back to Moodle, which I think a simply wonderful program from its basic functionality to its look and feel, and including its documentation, too. Though I can't say with any precision how it actually relates to "constructivist pedagogy", I think it helpful to ask how we might best relate the technology to the relevant, local pedagogies. As I've begun to outlined above, I've been looking for the points of access or entry. Many in my institution treat the technology as essentially an administrative problems, a matter of "secretaries, typewriters and toilet paper" is how I put it, and I think the more relevant and ultimately successful foundation will be on the research university's most fundamental terms: research, the seminar, and the preparation of researchers. I do not these days speak of teaching or pedagogy because, though there are many concerned with teaching and who are good at it, the institutional logic pays only lip service to it: no professorial review, tenure, and remunerations policies and decisions that I know of here are in any way based on good teaching.
I know not everyone using Dotlrn is in a research university, never mind a German research university, but I think that I am getting at some fundamental issues of user-oriented development, of which documentation is an essential part.
The idea of our seminar is that we build a syllabus, tutorials, and cohort of students and faculty/staff that includes a dozen powerful uses of the technology that are immediately relevant, that deliver a positive return on investment, and that Dotlrn use is one essential piece of a sophisticated academic toolkit that includes an expansive conception of email use (including association with personal databases such as Zoot), search machines and academic database use, group and communications functions, etc.
I must repeat one of the many lessons I learned from Al Essa's presentation and discussions at the Online Educa conference: that the case for Dotlrn can best be built on solid business terms (rather than a comparison of features or open source ideology). The business of the research university (I know, not everyone using Dotlrn is in a research university) is research: research is greatly assisted in many different ways with many different web technologies: Dotlrn delivers one or more specific features among the many needed with a better price/performance ratio.
Thus, my approach to the problem of Dotlrn documentation is part of a larger problem of professional use, and I would design and the help system in terms of that use. In these terms, I think that, better than Moodle, the better model for us is the BaseCamp project management software: their website explains the technology use in terms of the professional business use AND sells it to you in a bright, attractive package at the same time.