This is a request to gauge the real world impact of user agents and their capabilities / access scenarios in order to determine browser support requirements in forthcoming work.
We would like to be able to understand the counts of pageviews and completed edits from web browsers with an indicator for each of the following for each record.
- noscript disposition (in the context of completed edit, <textarea> edit disposition is also additionally instructive)
- grade A/X or C disposition (implicitly null if noscript)
- feature detector bitmask or map (implicitly null if noscript). In the initial launch, would like to know ServiceWorker capability. Here specifically, we don't want to exclude JS-supported browser from being feature detected.
- spider or user disposition
- proxy disposition
Preferable would be
(1) ongoing timeseries visualization:
- Grafana dashboard for global trend
- Grafana dashboard for global desktop web trend
- Grafana dashboard for global mobile web trend
- Grafana dashboard for negotiable list of countries x (desktop | mobile). (Assumption: de-proxied origin country is always captured.)
- there are meaningful tuples here, would be worth discussing if can visualize them in timeseries in some comprehensible fashion
(2) ongoing Hive queryable data. The Hive queryable data should be able to roll up namespace, project, and logged in disposition to understand ramifications for different user segments. Datetime information is assumed. A number of the additional fields of the nature of those in NavigationTiming and webrequest would be handy for deeper data analysis, most notably: deviceMemory, netinfoDownlink and referer_class.
The challange of fusing together data at multiple ingestion points dependent on client capabilities is very much appreciated! In some ways it seems almost theoretically possible to arrive at answers with existing tables, but upon further reflection, explicit instrumentation with ongoing visualization / data store would be more reliable.