Bug 1205567 et al showed that subsessionLength behaves inconsistently across platforms: Currently, depending on the platform, it may increase over sleep times or not. This poses a potential problem for analysis across platforms and comparisons against v4. I'm worried about potential implications for the v2/v4 exec dashboard sourcing - v2 is based on totalTime, per mreid v4 is currently based on accumulated subsessionLength. We might want to consider to base the v4 extract on totalTime to make them comparable (caveat here though: totalTime is time since the beginning of the session, not subsession - accumulation over subsessions won't work).

Heads-ups.

(In reply to Georg Fritzsche [:gfritzsche] from comment #1) > Heads-ups. Saptarshi, who owns decisions such as this one that affect the content of the dashboard?

Ultimately it depends on John. FWIW, as Georg pointed - V2 is totalTime - V4 is subsessionLength In any potential discrepancy, definitions of rollup measures based on V4 _must_ conform to definitions based on V2. So assuming totalTime is measured and means the same in both V2 and V4, the V4 rollups ought to use totalTime. In v2, totalTime is time of session.

IIRC we've discussed this before and made the decision to use totalTime from v4 for the executive dashboard.

totalTime isn't documented at https://gecko.readthedocs.org/en/latest/toolkit/components/telemetry/telemetry/main-ping.html Since totalTime is based on the entire sesssion and not the subsession, I don't know how we could create the proper rollups with the latency that we expect. For comparison purposes it's fine to create a v4 rollup which matches v2, but ultimately what we care about is not what we've done in the past but what is actually correct. I will express a requirement that whatever metric we end up with needs to be subsession-based. So if we're going to use something like totalTime, it needs to be broken down by subsession. And I recommend that we exclude sleep time from the calculation, but ultimately that is a decision that either Nick Nguyen or Rebecca needs to make.

RESOLVED in a meeting today. We're going to use subsessionlength right now. Longer-term we want a consistent number across platforms. For product metrics (especially things like crashes/time) we want a subsession length that excludes sleep time. Georg if there is not a bug on this specifically please file one. Nick or another product manager will help decide which other numbers are required in order to measure usage the way they need to see it.

(In reply to Benjamin Smedberg [:bsmedberg] from comment #6) > Longer-term we want a consistent number across platforms. For product > metrics (especially things like crashes/time) we want a subsession length > that excludes sleep time. Georg if there is not a bug on this specifically > please file one. * bug 1205567 is about making sessionLength & subsessionLength consistently include sleep time * bug 1205985 is about measurement for "time Firefox was open and computer not sleeping" Please follow up on the discussion in bug 1205567 if needed.