I would like to suggest the possibility that a reference manual 
slavishly integrated to the structure of Classes and methods may be too 
much and not enough.  I echo the sentiment that I'd like to see 
annotations of that structure reflected back in the image as class and 
method comments (or perhaps comment methods where more elaborate 
discussion is desired).  However, I have serious doubts that a 
comprehensive reference manual by extrapolating all that information in 
printed form will be either useful or helpful.  I fear it may be too 
much and not enough.

For the record, all the tables Squeak Quick Reference were built 
automatically with some workspace code, and then massaged at length.  
The workspace code intentionally eliminated redefined methods in 
subclasses, and made some other simplifying assumptions.  But I 
discovered that the document was useless pedagogically until I pruned a 
large number of obscure or redundant methods.

Clearly, this is a tricky business.  Prune too much, and the document 
becomes a mere primer.  Prune too little, and the reference becomes 
virtually indistinguishable from a printOut -- a static substitute for 
the browser.

The point is that not all classes are important for individual 
discussion, and not all methods are important for individual 
discussion -- at least from a pedagogical point of view.  Further, that 
there are times when class and method structure is less important than 
free-form prose with some clear examples.

The most important thing a teacher can do, I believe, is identify where 
chunks of information should be combined and summarized, and where a 
detailed analysis is truly profitable.  A good reference should do 
likewise.