Re: [whatwg/dom] High resolution timing for events (#23)

It was a good chat indeed. :)

I am trying to see if I can gather more data using telemetry to better help make progress on the debate about the potential impact. This should help address some cases where our earlier [static analysis of httparchive](https://github.com/whatwg/dom/issues/23#issuecomment-181518064) would have missed. I can only do this mainly thanks to @RByers recent efforts to enable cluster telemetry for Blink that allows such measurement for a set of top sites (e.g, top 10K).

Following @wycats suggestions I think we should measure the following cases:
  1. anyDate - timestamp  or timestamp - anyDate
  2. new Date(timestamp)
  3. date.setTime(timestamp)

Where timestamp is a value that came from Event.timeStamp **

As pointed out earlier, this will over count because there is no guarantee that any of the operations above will actually lead to breakage but we can followup with a manual check on this smaller subset. I am discussing with our V8 experts to see if this is actually possible without a huge amount of work. I will update once I have an answer.

In any case, before spending too much effort on this I like to have some agreement that this is an acceptable experiments and an acceptable threshold (e.g., 0.02% of pages?) for seeing this type of usage leading to breakage.

Also particularly interested on what @rniwa think of the effectiveness of this measurements given his strong concerns on web-compat.

** I have not yet figured exactly how to tag timestamp and Date values as I believe they get converted into a numbers before being operated on. Perhaps I will use a special value (may impact program flow) or specific fractions (A lot less intrusive)

---
You are receiving this because you are subscribed to this thread.
Reply to this email directly or view it on GitHub:
https://github.com/whatwg/dom/issues/23#issuecomment-215855675