> Could you give some hints of why this stuff is needed? It can't be in > order to merely keep the references right, because that could be solved by > avoiding having multiple aux files. To get references right means that the page breaks need to be the same even if some sections are not included. That means keeping track of where the floats would have floated to had they been included. Even if there are no floats, you need to know how far down the page each section should start so the page breaks work out the same again. \pageref is a harder beast to deal with than \ref, but \include doesn't just aim to keep references correct, but to keep all the typesetting of the included sections correct. So there is work to do even if there are no floats and no cross references. You need to know where on the page each section starts if you give up forcing that it starts at the top of the page. > I have not followed that gory float stuff, but this is what I am saying: > Are there any reasons for LaTeX having multiple AUX files? Yes there are several good reasons. It is (perhaps) not impossible to avoid the use of multiple aux files but it is not entirely trivial. In particular if a section is not included then currently its .aux file hangs around for use in later runs. If the information for the section is written to the main aux file then it will need to be copied if that section is omitted, but how to copy from the old aux file to the new one? either you store the full list in memory or you have to have two aux files and keep switching from one to the other, or... Similarly the separate aux files store the counter resetting information to use when a section is skipped. Being in separate files means the .aux file can be input at the appropriate time (in place of the skipped file). If all the information is in one file either that file has to be read multiple times, or all the information needs to be held in memory so the counters can be reset when needed. However Why do people use \include? Some possible answers 1) By mistake, when they meant to use \input. 2) To speed up processing of drafts as only the `current chapter' need be processed. 3) To cope with big jobs where running the whole document in one go runs out of memory. I think (1) is probably the most common reason. (2) is not relevant if you are using a class where the individual sections are self contained documents. To just work on section1 you don't need to go \includeonly{section1} you can just go latex section1.tex So that leaves (3). But (3) is usually a forlorn hope. Often the reason for running out of memory on big jobs are too many \labels. As \include always reads all the \label info from the aux files, it does not help here at all. The only time it can help if the separate chapters use different large packages, eg if section1 uses pictex and section2 uses xypic, you may gain a bit by using \includeonly to process the sections in separate jobs. But here again probably you usually run out of register allocations before you run out of csnames, and that can be fixed by a better allocation scheme (or fixing pictex not to waste so many registers). David