We want to know how much bandwidth SafeBrowsing is using, to make an informed choice on when to enable it, and probably the users also want to know this. The simplest (and most accurate way) to me looks like running a browser on a dedicated machine and sniffing off all traffic from that box to the Google update service. This allows to make a distinction between the initial database download (day 1) and normal usage (following days). It also avoids complications when the browser requests an update after a restart (it'll tend to be tiny, which complicates gathering statistics over telemetry).

11.3Mb used for the first day - this includes the initial sync of a fresh database. https://docs.google.com/spreadsheet/ccc?key=0AiObroK0ZHF-dEVoa0VINHFFVndxZDZSSmRsR3BsalE It's also in line with the previous measurement that the database contains about 9M of uncompressed data. https://docs.google.com/spreadsheet/ccc?key=0AiObroK0ZHF-dGFmZWhKLVdSWnhYcGk1VDZaZjg3X1E Will have data on daily churn in a few days.

Churn traffic on day 2: 860kB.

not blocking fennec. this probably should go into Core / <some component>

Traffic up to day 5 added. First day: about 11M used. Subsequent days use about 900k/day. I think this is enough data.

By the way, changing the system to do something like "only update on wifi" is really difficult because the protocol has hard requirements on being up to date: http://code.google.com/apis/safebrowsing/developers_guide_v2.html "Applications retrieving data using the API must be certain never to use data older than 45 minutes." So it's either On or Off.