Copy from a email from holte@ on what we want:
As an example of what this might look like in our data and how we would use it, we could consider a user with the following timeline of actions:
* Open normal Chrome on "http://startpage.com"
* Open an incognito window on "http://incognito.com/"
* Open a normal tab to "http://otherpage.com"
* Close the incognito tab
* Do a contextual search on "http://otherpage.com"
This could currently produce a report like:
{client_id: 123, report_id: 1,
sources:[{id: 1 url:"http://startpage.com"}]
events:[
{source:1 hash:Pageload},
{source:3 hash:ContextualSearch}
]}
}
And this report is odd, because we have an event associated with a source we don't know the URL of, because the URL recording occurred while recording was disabled for incognito. There are other possible explanations that are currently possible, like source:3 could be on some non-https page, some bug in data collection, a spam report. It would be useful if we could annotate the report that we expect some data discontinuity e.g.:
{client_id: 123, report_id: 1, has_discontinuity: true
sources:[{id: 1 url:"http://startpage.com"}]
events:[
{source:1 hash:Pageload},
{source:3 hash:ContextualSearch}
]}
}
So that we know that some missing data is expected.