... historians should cite the text they are using, and not invent virtual archives that lead one to believe that they are using the manuscript text in an archives, and not the text presented in a documentary edition.
So, if a historian is citing a letter of William T. Sherman to someone that appears in Sherman's Civil War, and that's the text they are using (and not, say, the Sherman Library of Congress collection, whether it be the microfilm edition or the actual document), they should cite my text, and make it clear that such is the text they are using.
If I was citing a letter that appeared in John Y. Simon's documentary edition of Grant's papers, and it was that text (and not something I found elsewhere) I was citing, I'd cite Simon. When I don't cite Simon, you can rest assured I've dealt with the original document. (You can also rest assured that I have dealt with it, because I don't use research assistants.)
Let's not pretend...
Speaking of citation, another historian, Brooks Simpson, shared with me this principle of practice.