Some other ideas include: - I/O lists. You can submit atomic I/O lists to be executed by the kernel or by a proxy, including locking and transactions, and atom

Some other ideas include: - I/O lists. You can submit atomic I/O lists to be executed by the kernel or by a proxy, including locking and transactions, and atomic I/O with locking and transactions is also possible on multiple files at once. - You can duplicate the running state of the system (or of parts of it), to be able to restore it later, examine it for debugging, etc. This is not always possible (e.g. due to hardware states), but would try to usually be possible. - A program's configuration file normally would link to the program file, or else another file links to both of them; that is the program that you must call in order to start the program (normally), rather than the program itself. In this way, the program can find its configuration files and other files. If you do not want a program to update its own configuration file, then you can just make that file read-only, or use the versioning capability of the operating system. (These still aren't all of them, though)