Not sure how high of a priority this is, but I don't want it to be lost: We're not including "system" network traffic (appId==0) in our network stats. We should fix that, as mentioned in bug 887699 comment 33: > WebSocketChannel::SaveNetworkStats(bool enforce) > { > #ifdef MOZ_WIDGET_GONK > // Check if the connection type and app id are valid. > + if(!mNetworkInterface || mAppId == NECKO_NO_APP_ID) { So here and in the HTTP patches too, you're not recording traffic with id=NO_APP. But this includes any networking the phone does that's not from an app (like updates, getting phishing lists, etc.). Since your nsIDOMMozNetworkStatsManager.getSample method says "If appURL is provided, Find method will retrieve per-app usage, otherwise the target will be system usage", I'm guessing you want to actually keep track of NO_APP traffic. It will be a more accurate measure of actual system data traffic if you do.

(In reply to Jason Duell (:jduell) from comment #0) > Not sure how high of a priority this is, but I don't want it to be lost: > > We're not including "system" network traffic (appId==0) in our network > stats. We should fix that, as mentioned in bug 887699 comment 33: > > > WebSocketChannel::SaveNetworkStats(bool enforce) > > { > > #ifdef MOZ_WIDGET_GONK > > // Check if the connection type and app id are valid. > > + if(!mNetworkInterface || mAppId == NECKO_NO_APP_ID) { > > So here and in the HTTP patches too, you're not recording traffic with > id=NO_APP. But this includes any networking the phone does that's not from > an app (like updates, getting phishing lists, etc.). Since your > nsIDOMMozNetworkStatsManager.getSample method says "If appURL is provided, > Find method will retrieve per-app usage, otherwise the target will be system > usage", I'm guessing you want to actually keep track of NO_APP traffic. It > will be a more accurate measure of actual system data traffic if you do. Hi Jason, I've already filed a bug about this in bug 922926. In the beginning, "system usage" (refers to the total usage that obtain from netd) was just used to distinguish with the "per-app usage". But what you point out is also important because we didn't consider about the "system-only usage", so we will fix it. Since the bug is already filed, I'm closing this one hope you don't mind. Thanks