Each page could not contain it's own title tag because one document could not contain more than one. You would just have to get it from the filename or something.
I was thinking that the entire "book" would be split into chapters, each chapter in a separate HTML file. Each HTML file would be a well-formatted full-fledged HTML, with the header and everything. Then there could be an optional external tool that would take these files, generate a TOC HTML file for them, and optionally one huge HTML file with the contents of each file's body.
Am I overcomplicating things?
The only problem is that if you want to change the header, you have to go and change 20+ files to change the header. And if you're changing the header 10 times a day, that's going to get annoying.
A templating system would fix that. If we use github pages jekyll is the obvious choice.
We would not be using GitHub Pages.
What I really should have said is that I don't think jekyll is a good idea - especially if we want to generate one megafile with all the sections in it.
Jekyll can handle that, just have a file that includes all the sections. If you don't want to use jekyll can I recommend
Template Toolkit Available in perl and python versions; or
handlebars.
I've used basic python templates before and they worked perfectly fine. I'll see about getting a basic generator up in Lua or python, it shouldn't take too long.
I've tried writing a template engine before. It is seriously worth using the existing template engine as they provide loads of extra features.