Dave Miner wrote: > I'm skeptical that the above questions actually represent useful input > to a heuristic engine. "Both" seems like it'll be true way too often to > #1, plus I don't want to have to remember to change it when I decide I'm > going to go from a pure client to add an NFS server. Really, it seems > like we could figure this one out for ourselves based on what services > the user chooses to enable. For #2, there's a lot of modality buried - > my laptop might sit on my desk at home, changing rarely, except when I > go on a trip, when it goes into "often", and again, I don't want to have > to remember to make that change, either.
Could the number of configured network interfaces be one determining factor in (auto-)configuring the user intent knob? If there is at most one non-loopback interface of the same type, the system is configured as a "desktop" or "end user" system, if there are more than one, its a "server" or "enterprise" system? What other things could we factor into making the decision automatically? Presence of certain "high end" hardware, amount of memory and so forth? Obviously, none of these represent things that strictly can't be found in either configuration. Also, to make the knob more useful as an install-time question, could it potentially be tied to deciding what bits get installed on the system? (similar to the Linux distros install time choices of "Gnome Desktop", "Minimal" and "Everything".) The user intent knob may also allow simplifying the network service model- an end-user system may take more liberties with automatically refreshing services based on network context changes. -Anay
