“Optimising” document formatting is an iterative process. Unless you are exceptionally smart (when it comes to aesthetics issues, I am certainly not), you never design your “ideal” styles from the start. The most important point is to define the set of semantic styles you need. I write semantic because you must consider styling as a “significance annotation” of your text. As an author, you, and only you, know which paragraphs or words are headings, main topic, comments, notes, important, humour, irony, … The name you give to the corresponding styles won’t change during the course of writing. Don’t bother yourself with the appearance at this stage. As long as the styles (their names) allow to make a distinction between the “values” in the document (the abstract significance categories), you can defer the formatting tuning step. Just ensure the present state of the styles give you a sufficient visual feedback.
When it is time to tune formatting, play only with the styles. If there is no direct formatting, there is nothing else to do. This is why it is important to reduce direct formatting to the absolute minimum. Otherwise it is a real nightmare. You’ll soon discover that, contrary to common belief, use styles is much simpler, efficient and effective than direct formatting.
One last word: don’t go for master+subdocuments unless there is a real need for it. Present-day computers are powerful enough to handle 800-1000 pages single-file documents (without images or tables), even laptops. The limit may be smaller if your document structure is complex.
The master feature is a good solution when the sub-documents are shared and reused without modification in several “works” (books). But if the document is unique, there is no real reason to make things more difficult unless you’re on several thousands page book. The master feature introduces subtleties with new difficulties. Consequently, think twice before deciding.
