We should improve our current telemetry to extract meaningful information about:
advanced chart configurations (analze precisely the usages of each chart type and feature)
forecast disruptions of feature deprecations
remove/reduce unused features and tools
understand used feature workarounds
To achieve that I think we should get more details about the current chart configurations, rather then just counting the usages of each single feature.
Chart configurations are mostly non-sensible information that can be parsed and used to analyze the general trend and usages of features.
The only possible privacy-sensible parts of a chart configuration are:
titles and descriptions
field name/labels
axis name/labels
markdown content
indices names
The gathering method could be done on a demand basis, without any particular schedule.
In the future, this gathering could be done systematically in a scheduled manner to understand the usage trend over time.
We should improve our current telemetry to extract meaningful information about:
To achieve that I think we should get more details about the current chart configurations, rather then just counting the usages of each single feature. Chart configurations are mostly non-sensible information that can be parsed and used to analyze the general trend and usages of features.
The only possible privacy-sensible parts of a chart configuration are:
The gathering method could be done on a demand basis, without any particular schedule. In the future, this gathering could be done systematically in a scheduled manner to understand the usage trend over time.