Hi Rick,
Many of the ideas that are discussed in the "Names" chapter will be expanded upon in the chapter on "Indirection", namespaces included. With that said, I did try to discuss both organizing by data scope (or layer, as you say) or datatype. I suggest avoiding the "one namespace per datatype" approach only where there are a large number of datatypes, because we don't want to have too many namespaces. It seems like in your case, it's worth having a few more. I'll think about how I can be more clear about this in the first chapter, and how I should expand on it in the third.
As to the rest, I'm not wild about the "complection" neologism. Putting aside Rich's definition in the talk, I feel like it's used to describe two separate problems:
* treating two distinct ideas as the same (which is adequately captured by "conflate")
* implementing two clearly distinct ideas in a way where they're not easily separated
Obviously the first problem can lead to the second one, but I think they are different enough that they should be talked about separately. Using the terminology in the book, I think it's more clear to talk about the second problem as "insufficient indirection". Again, though, I'll give it some thought. Nothing's set in stone.
Thanks,
Zach