Nothing yet, unfortunately. There is discussion around a “FetchObserver” or another mechanism to listen for the “start” of resources.

There’s some discussion here: https://github.com/w3c/resource-timing/issues/137 and please include your feedback as it can help push issues forward!

Quick question: Is there a way to get the list of resources that are “in progress”?

I’ve notices that the performance.getEntries() array populates when the transfer is done, but does not give any hints about the resources that are in-progress.

@Raffi: Yup! See the Timing-Allow-Origin section: https://nicj.net/resourcetiming-in-practice/#timing-allow-origin

There is a header — Timing-Allow-Origin — that can be set to have the values reported correctly (have not tested myself but seems likely.)

when we get ‘use’ initiator type

URL used on mobile site (PWA) – https://in.bookmyshow.com/buytickets/parmanu-the-story-of-pokhran-mumbai/movie-mumbai-ET00058781-MT/20180606

There is an open Chrome bug on the blank initiatorType: https://bugs.chromium.org/p/chromium/issues/detail?id=649571

You could add a comment, or star it, to let them know it’s affecting you. In the meantime, you could UA sniff Chrome and if the initiatorType is blank, there’s a high liklihood that it is fetch — I’m not aware of any other APIs that have a blank initiatorType today.

Hi Odi,

ResourceTiming data is completely separate from things like XMLHttpRequest because it was designed to holistically report on everything in the page, instead of having direct tie-ins to each component.

That being said, you may be interested in the discussions on adding a “fetch ID” or “request ID” that might help correlate specific XHRs (fetches) with ResourceTiming entries:

https://github.com/w3c/resource-timing/issues/8
https://github.com/w3c/resource-timing/issues/39

@Mike: Hi Nic, could you please tell me a little bit more about on what is “all redirected resources” ? I noticed that most of the time, duration = responseEnd – fetchStart for regular image/JS resources.

If you’re downloading http://a.com/foo.png and it 301 redirects to http://a.com/bar.png, the time it took to do the redirect will be between startTime and fetchStart.

@Mike: If I can guarantee that the script will not be redirected for sure then we are safe to rely on the “duration” value for cross domain resources right ?

I think so, yes.

Thanks for the great article!
For completeness, I would add Dynatrace UEM on top of New Relic and AppDynamics offerings, as it captures all W3C Resource timings for every action of every user.

And yes I work for Dynatrace. However I thought if you add 2 then you might just as well ad the third as well ;-).

Cheers,
Kristof

If I can guarantee that the script will not be redirected for sure then we are safe to rely on the “duration” value for cross domain resources right ?

Hi Nic, could you please tell me a little bit more about on what is “all redirected resources” ? I noticed that most of the time, duration = responseEnd – fetchStart for regular image/JS resources.

Thanks in advance!

Thank you Nic for the reply. I will take a look at plugin you mentioned.

Thanks for the research! I’ve updated with your comments. Bummed I didn’t bump into you at VelocityConf EU!

I’m still seeing duration=0 for FF 43 on Windows for DNS failures. Strangely, on Mac FF 43, I don’t see any ResourceTiming entries for DNS failures via Andy Davies’ test case: http://andydavies.github.io/rt-tests/dns-failure.html

ResourceTiming only notifies you of resource fetches after they are complete. In order to instrument all XHR activity, you may need to overwrite the XMLHttpRequest object. See auto_xhr.js in Boomerang for one way of doing this: https://github.com/lognormal/boomerang/blob/master/plugins/auto_xhr.js

how can I use Resource Timing API to instrument XmlHttpRequests and log timing as soon as that happens? Just curious what event is fired before XHR resource starts and ends similar to Network tab of Chrome Dev tools? Please advise. Thanks a lot

More people should write these kind of deep-dive articles. Well done!

Some feedback based on tests on FF 41.0.1 (latest) on Mac OS X Yosemite (10.10.4) :

DNS Failure: “domainLookupStart through responseStart and duration are 0. responseEnd is non-zero.”
I consistently see duration being non-zero, and seems to be the actual time it took, eg. 282.43245599999955.

TCP Failure: “domainLookupStart through responseStart and duration are 0. responseEnd is non-zero.”
This is still the case, which is cool, because now it can be distinguished from a DNS Failure.

404: “All timestamps are non-zero, though duration is 0.”
I see duration being non-zero and the value ‘looks about right’.

PS: coming to Velocityconf EU 2015 in Amsterdam? If so, let’s meet for coffee and RT API chatting.

Great catch. Apologies for not actually testing the code I was showing!

It looks like Chrome supports onwebkitresourcetimingbufferfull and FireFox supports onresourcetimingbufferfull Functions on the performance object. This may change with the latest draft as it’s supposed to be an EventListener now (http://www.w3.org/TR/resource-timing/). IE doesn’t appear to support either, I’ve pinged them to see if we’re missing something.

I’ve updated the sample code above.

Are you sure that document.addEventListener(“onresourcetimingbufferfull”, function { // do something }); works?
The onresourcetimingbufferfull event is inside the window.performance object. Moreover, IE 11 does not even have this event to attach to. Any idea how to overcome this?

Any idea on why we need ALPN Protocol Id on the RT? what is the use case its solving?