Will browsers eventually let a site request a HAR file from a regular user?

132 views
Skip to first unread message

Chase

unread,
Apr 15, 2011, 4:06:23 PM4/15/11
to HTTP Archive Specification
We have been researching various client-side performance measurement
tools, but it seems that none of the existing tools can get a complete
picture of client-side performance. Tools like Jiffy measure render
times in javascript, and cannot account for time spent before the page
first comes back. Tools like yottaa and browsermob automate browsers
around the world, but cannot account for real user data.

It seems like the perfect tool would be for the browsers themselves to
record HAR data, and then upload it to the web server on request,
maybe triggered by a custom HTTP header.

Are we headed in that direction?

Guy Podjarny

unread,
Apr 18, 2011, 7:27:28 AM4/18/11
to http-archive-...@googlegroups.com
This sounds like a great enhancement, but I think it won't be able to
capture real user data.

In order for it to capture real user data it would have to allow
JavaScript to access the HAR files and upload them, which has major
security and privacy implications.

So I think the end result would be a capability like yslow's
measurement, meaning you grab the browser's timings, but only when
it's your own browser.

One simpler bit of info browsers could make available is the time the
page request was issued. This would make JavaScript based timing
measurements more accurate, without compromising security.

Cheers,
Guypo

Guy Podjarny | CTO, Blaze | 613-800-0413 x202

Pat Meenan

unread,
Apr 18, 2011, 8:14:20 AM4/18/11
to http-archive-...@googlegroups.com
Actually, that's exactly what the Resource Timing part of the web timing
spec is trying to tackle. It's probably still a ways off from being
sorted out because it's a pretty nasty hairball of security and privacy
concerns but the goal is to give javascript on the page access to
per-resource timing information -
http://lists.w3.org/Archives/Public/public-web-perf/2010Nov/0001.html

As far as the start time (and actually a lot more great data) is
available from the web performance spec which is already implemented in
Chrome and IE9 (with Firefox support coming later this year) :
https://dvcs.w3.org/hg/webperf/raw-file/tip/specs/NavigationTiming/Overview.html

Lots of great work going on in the browsers to be able to get real field
performance data.

Chase

unread,
Apr 18, 2011, 10:25:56 AM4/18/11
to HTTP Archive Specification
Wow, I had no idea the basics were already available in IE9/Chrome. It
seems like that's enough to start using this right now.

Has anyone heard of a cloud hosted vendor that will aggregate this
information via an Ajax call from a javascript include? I'm surprised
Google Analytics doesn't have this already...

-Chase


On Apr 18, 8:14 am, Pat Meenan <patmee...@gmail.com> wrote:
> Actually, that's exactly what the Resource Timing part of the web timing
> spec is trying to tackle.  It's probably still a ways off from being
> sorted out because it's a pretty nasty hairball of security and privacy
> concerns but the goal is to give javascript on the page access to
> per-resource timing information -http://lists.w3.org/Archives/Public/public-web-perf/2010Nov/0001.html
>
> As far as the start time (and actually a lot more great data) is
> available from the web performance spec which is already implemented in
> Chrome and IE9 (with Firefox support coming later this year) :https://dvcs.w3.org/hg/webperf/raw-file/tip/specs/NavigationTiming/Ov...
>
> Lots of great work going on in the browsers to be able to get real field
> performance data.
>
> On 4/18/2011 7:27 AM, Guy Podjarny wrote:
>
>
>
>
>
>
>
> > This sounds like a great enhancement, but I think it won't be able to
> > capture real user data.
>
> > In order for it to capture real user data it would have to allow
> > JavaScript to access the HAR files and upload them, which has major
> > security and privacy implications.
>
> > So I think the end result would be a capability like yslow's
> > measurement, meaning you grab the browser's timings, but only when
> > it's your own browser.
>
> > One simpler bit of info browsers could make available is the time the
> > page request was issued. This would make JavaScript based timing
> > measurements more accurate, without compromising security.
>
> > Cheers,
> > Guypo
>
> > Guy Podjarny | CTO, Blaze | 613-800-0413 x202
>
Reply all
Reply to author
Forward
0 new messages