Just to be contraversal, I'm going to suggest 'The Old New Thing', by Raymond Chen.
Not so much for the details of windows (although those are interesting), but it is the only book I have ever seen that attempts to show how a large software project evolves over a long period of time.
Not so much for the details of windows (although those are interesting), but it is the only book I have ever seen that attempts to show how a large software project evolves over a long period of time.