For writing a PhD thesis in LaTeX, there are many package choices to make and a whole lot of little issues and pitfalls to handle. This starts with choosing the document class and … goes on and on. Here, I am collecting some points which took me more than a few moments, so I don’t have to go through the whole process again. And maybe my little list can help others to make their choices a little faster or avoid some of the mistakes I made.
Choosing the document class
Even though there are alternatives, for me, the main contenders for the document class were KOMA-Script class ( scrbook or scrreprt ) and memoir. In the end I chose KOMA-Script as it seems to be less bloated (i.e., memoir emulates many well established packages instead of using them directly) and there really seems to be no reason not to use it 😉 However, as you can see, I am not going into minute details as the articles mentioned above were enough to sway me.
As for the subcategory I use scrbook . It seems that from the book, report or article classes (koma-script or not) there seems to be an awful lot of features which fit a PhD thesis more than the report or article class (see this discussion). Just to name a few, such features include two-page layout or more flexible numbering with regard to front, main and back matter. Also this article explicitly states that scrbook is for dissertations.
How to handle subfigures
An interesting choice was which package to use for subfigures. I chose the subcaption package. As to why: the subfigure package seems to be old and is superseded by subfig. While the newest most flexible option seems to be a combination of the caption and the floatrow package. However, the subcaption package supposedly was a case study for the interfaces provided by the latter two packages and provides a more easy to use functionality. Indeed, the introduction of the subfigure package on CTAN specifically describes itself as “superseded […], […] and users may find the more recent subcaption package more satisfactory.” Discussions on StackExchange further support this notion. I think overall, it boils down to subfig supposedly having issues with hyperref, and subcaption being more simple than using a combination of caption and floatrow. Thus in the end subcaption seems to be the best choice.