>>>>> "D" == David Carlisle <[log in to unmask]> writes: D> 2) To speed up D> processing of drafts as only the `current chapter' need be D> processed. D> (2) is not relevant if you are using a class where the D> individual sections are self contained documents. To just work D> on section1 you don't need to go \includeonly{section1} you can D> just go latex section1.tex Hunh? Like what kind of class is that? Do I misunderstand what you mean? You can't just go latex section1.tex if it doesn't have a \documentclass in it. If it does, either you weren't using \input or \include to include it in the first place, you were using something like \includex; or you are using a sophisticated LaTeX front-end like AUCTeX which supplies a header and footer for section1.tex from somewhere else. D> processed. 3) To cope with big jobs where running the whole D> document in one go runs out of memory. D> So that leaves (3). But (3) is usually a forlorn hope. Often I agree. The only time I have ever had TeX memory size problems is doing intensive fiddling with pstricks to cross-shade a large table. A typical 300+-page document (source2e for example) doesn't tax TeX at all, and we are about to enter the era of web2c-7 with dynamic memory allocation. When I tried the solution of one aux file and saving all the parts' auxinfo in memory with macros (tag.sto, group.sto). I did some experimentation and concluded that the extra memory usage was irrelevant. I don't say this would be so in all applications, of course, but the result was encouraging.