On Wednesday 09 March 2005 12:29 pm, Jordan Crouse wrote: > On Wed, 9 Mar 2005 10:52:24 -0800 > "David Brownell" <david-b@xxxxxxxxxxx> wrote: > > > There's something to be said for an integrated policy ... but at > > the same time, cramming too much stuff into the bag can leave you > > with a lot of holes ripped in it by sharp corners sticking out! > > There are times it's better to use more than one bag. :) > > I like that analogy. It worked there, but I'm not sure how reusable it is! > You make some great points. I just want to avoid a situation where we > have power thresholds spread out all over the sysfs, which in my eyes > takes a fairly complicated and intelligent user agent to deal with, > especially when one would need to manage large numbers of devices at a > time. Far better to offload that complexity to the user, and force > them to craft an complete policy instead. I'm not sure what you mean by "thresholds"; can you elaborate? That notion hasn't come up here recently. I'd think that for example a battery manager would care about thresholds as one set of event sources ... especially for non-"smart" batteries. > Let me qualify that I'm coming at this from a completely > embedded perspective. In my mind, I'm power managing a nice closed > device with a fixed list of devices. I realize the real world doesn't > always work like that. If it's a fixed list, then there's also a fixed list of sysfs files that would be configured in a policy. So it won't matter whether or not they're spread out "all over". When the list includes things like USB hosts (possibly through OTG) or CardBus bridges, it's a bit trickier but you'll at least have rules about where new devices will be added (relative to the existing ones). > Anyway, this is why I mentioned DPM, because the general design nicely > mirrored what my concerns and requirements were. Yes, it's certainly worth a look by folk who haven't yet done so. - Dave
