I like to think of myself as an article writer who writes the occasional book. With my dissertation book, I did things the standard way: I wrote the manuscript, which upon completion of review I submitted (as LaTeX and supporting files) to Cambridge University Press for production. It was a hard experience. Numerous errors crept in during the typesetting process, which a kind employee at Cambridge and I spent a week cleaning up. (I think we caught them all.) A typical error: \frac{a}{a+b} rendered as a/a+b. Ugh.

Second time around, when I wrote my textbook, I did the typesetting myself. My primary motivation in doing so was to ensure the printed accuracy of what I had written, but I had also noticed that some texts were set better than others, and I wanted some control over the process. I knew quite a bit of LaTeX at this point (if you feel more comfortable in Scientific WorkPlace or LyX, this is not for you), and it was not difficult to do the typesetting, once I understood the process.

Memories of this flooded back recently after a conversation on Twitter with Jelena Subotic and Anna Grzymala-Busse. Much of my experience was idiosyncratic (a particular publisher, a certain type of manuscript), and some of it may be out of date. At the same time, some of these lessons (e.g., with respect to producing indexes) are likely useful even to those who are happy to let the press do most of the work.

Here, for anybody in a similar position, is what I learned. Caveat emptor.

Negotiate the process in advance with your publisher. With my textbook, we agreed that I would produce the proofs, the copy/production editor would mark up the proofs by hand, I would enter changes directly in LaTeX—this process iterated until we were both happy with the product. (In addition to language edits, my editor would suggest that I move a figure over a few pixels, etc.) I submitted a PDF file for printing. I also provided the LaTeX and supporting files, just in case I got hit by a bus and something needed to be changed, but they weren't needed and likely never opened. This all worked in part because I had an exceptionally good editor…if you have a very technical manuscript, you might ask if they keep a copy editor in reserve for that sort of work.

Don't reinvent the wheel: find the document class that your press uses. For me, it was cambridge7a. You will make many changes from the default set, but it's a place to start.

Ask the production company to produce the front matter (title and copyright page) and back matter (e.g., list of other books in a series).

As much LaTeX as you know, there will inevitably come a moment when you need to do something you don't know how to do. Cambridge has a support line at texline@cambridge.org. Other presses may have something similar.

You are saving the press money by doing the typesetting yourself. They may be willing to increase your royalties in return.

Finally, if you are enough of a perfectionist to consider typesetting a book yourself, you may also want to produce your own index. It is a good thing to consider: you know your book better than any professional indexer will, and you will save the time otherwise spent cleaning up nonsense entries. In LaTeX this primarily involves inserting \index commands at the appropriate place in the manuscript and then compiling. Think BibTeX: the process is similar. In my case, I also wanted an author index, which is truly automatic if you have used BibTeX for references, although it requires a Perl script and a bit of customization if you are working with natbib. Check out my notes for doing this on Windows and Mac OS: I switched from the former to the latter between my first and second book. There is almost certainly a cleaner way to do this—mine uses the deprecated package multiind, still available at CTAN. If you get lost…I don't know. I found most of this somewhere online.