Schema talk:NavigationTiming

From Meta, a Wikimedia project coordination wiki
Maintainer:Timo Tijhof
Team:Performance
Project:Extension:NavigationTiming
Status:active
Purge:Auto-purge just eventCapsule PII after 90 days, keep the rest indefinitely

NavigationTiming API[edit]

Abbreviate property names?[edit]

  • Should we abbreviate property names for the sake of minimizing the request size? i.e. dnsLookup: dL, connecting: c, sending: s, etc.
  • Should we replace userAgent with simplified buckets, i.e. ie/chrome/ff $majorversion?

--Asher Feldman (talk) 21:17, 12 March 2013 (UTC)[reply]

Yeah, the full user-agent string adds quite a bit of bloat, especially considering it's sent twice (once as part of the query string, and once as a standalone HTTP header). I'll take a look and see if there is a concise, cross-browser way to distill it. As for abbreviating property names, I think it will make sense to do that once the schema settles -- I expect that the initial deployments will show which measurements are actually interesting (i.e., track things that are within our power to change and that we don't already know about). So for the first couple of rounds, I say we keep things verbose. --Ori.livneh (talk) 08:14, 13 March 2013 (UTC)[reply]

Cf. mw:EventLogging/UserAgentSanitization.

Validating measurements[edit]

The W3C published a set of test suites for the Navigation Timing API to help browser vendors verify that their implementation conforms to the specification. The tests specify (in the form of assertions) the kind of relationships the various values ought to have with one another. It could be quite useful to take these tests and use them to verify the data we log. --Ori.livneh (talk) 05:39, 22 March 2013 (UTC)[reply]

Data[edit]

See NavigationTiming/Logging for data and analysis. Maryana (WMF) (talk) 00:05, 9 April 2013 (UTC)[reply]