JG> The way I see it, if I only use a number once, having it as a JG> define doesn't get me very far, because the define really just JG> creates a level of "indirection" for whomever decides the value JG> needs to be changed. As soon as the number is used twice, you of JG> course need a define so that nobody has to root out all instances JG> of the value, but for one-time-use-only values I don't generally JG> bother with defines. Well, even if something is only used once, it makes it easier to change something into a configure option if it already has that layer of indirection in place. It also means that someone can easily open the file, see that the limits are all defined at the top, and not bother to understand the rest of the function pointer indirection (which would be harder to follow without context than the #defines anyway). The only things that I think make sense to leave buried are things like increment for NICs, since it's very clear that adding one NIC will always be the desired granularity, which can't be said for disk or memory, IMHO. -- Dan Smith IBM Linux Technology Center Open Hypervisor Team email: danms@us.ibm.com