martin f krafft wrote: > while [ ! -e/var/run/my_semaphore ]; do sleep 1; done
I don't consider this a semaphore. A semaphore should avoid busy waiting. Using my implementation, busy waiting is avoided, and the script can go on immediately (when the task is scheduled), without the 1 second sleep. Exaggerating, using this kind of semaphores boot may become slower... >Don't get me wrong, I appreciate your effort. But before we add such >complexity (and TCP *is* complexity), we should know which problems >we are trying to solve therewith, which can't be solved with easier >methods. > If we would allow a quicker parallelized startup, we must add something. However I think we don't have to oblige admins to use parallelized startup. If someone want a boot simple and safe, he can use the old fashion method. If someone want a boot "quick and dirty", he can install a package and use this kind of startup. My method is backward compatible. Desktop users may appreciate this method, server admins may prefer the old method while this method is not mature. In any case, TCP is provided by the kernel through system calls, IMHO this doesn't add too much complexity. However, it may be replaced with local "unix" sockets, probably they are safer. _______________________________________________ initscripts-ng-devel mailing list [email protected] http://lists.alioth.debian.org/mailman/listinfo/initscripts-ng-devel
