Add web3dsurvey data gathering to specs and extensions.

[kenrussell]

Thanks to @greggman for suggesting the web3dsurvey refactoring to put
the data collector into a separate iframe.

Thanks to @kainino0x for helping me understand why the iframe was
breaking in the XSL-generated files, and for the workaround.

[kenrussell]

@bhouston did I do this correctly?

Note - if the script's inserted higher up in the extensions' HTML then it breaks their rendering. I'm concerned about this and wonder whether something is going wrong more deeply.

[greggman]

I want to suggest a few things

1. This script should not be a script, it should be a 3rd party iframe as in

   <iframe src="https://web3dsurvey.com/collector.html" style="width: 1px; height: 1px;"></iframe>
   

   or something along those lines. As a raw <script> it gives a 3rd party site permission to read cookies, local storage, etc of khronos.org or registry.khronos.org which might be storing login credentials.

2. if the goal of this script is to get data of the general population then is adding it to a site who's only audience is 3D devs adding or subtracting from that goal?

[bhouston]

This script should not be a script, it should be a 3rd party iframe as in

@greggman I will make the change you suggested - this is a good point.

if the goal of this script is to get data of the general population then is adding it to a site who's only audience is 3D devs adding or subtracting from that goal?

Interestingly, I can add correction for a biased samples. I will look into reporting on this as we go forward. Right now I am collecting about 100K samples per day. It is actually decent and while somewhat biased, it is pretty wide with lots of variation.

[bhouston]

@kenrussell just put the script in the body probably at the end - it shouldn't be in the section -- I will clarify that on the website. Then it should be fine. I will make a further update in a week or two that does what @greggman suggests and we could do a further update with that change - but I wouldn't wait of it.

[bhouston]

@greggman I was thinking about this and I may make it into two separate scripts. I will keep the collector.js script but it will be very small and just create an iframe that loads a second script. I will add a Subresource Integrity attribute to the collector.js so that when people embed it, they know exactly what they are getting. Although the seconds script that loads into the iframe will not have a content validation hash so I can continue to version it, but it will load into that sandboxed environment. Does that make sense @greggman ?

[greggman]

The advantage to the iframe is I don't have to verify your script. With Subresource Integrity, I need to go find the source of your script, read through it to make sure it's not doing anything fishy, verify that it generates the same hash. Then, if at some point in the future you need to update the script, I have to through that process again.

With the iframe, I don't have to do any work, just put the iframe in the page. The fact that it's a 3rd party domain is enough for me not to have to care what the script does

[bhouston]

I've updated the recommended embedding procedure on Web3DSurvey.com to suggest the iframe embed and I give example code on how to do that based on @greggman's feedback:

<iframe src="https://web3dsurvey.com/collector-iframe.html" style="width: 1px; height: 1px;"></iframe>

It is actually better for collection because on that collector-iframe.html page I can include an Origin Trial token to enable WebGPU on Chrome devices.

With this new approach @greggman would you be interested in embedding the script to help with data collection?

[kenrussell]

Apologies for forgetting to revisit this!

Thanks @greggman for your suggestion to use an iframe, and @bhouston for implementing that a while back. @bhouston , perhaps you would like to update your suggested code to use frameborder="0", which was necessary to keep the WebGL specs rendering cleanly.

Thanks @kainino0x for helping me understand why the XSL-generated HTML was breaking due to a self-closing iframe tag, which is illegal, and for the suggestion to use a non-breaking space inside the iframe to work around that.

I've verified in Chrome's Devtools that this works on all pages (the specs, extension registry, and individual extensions' pages), and that the collector runs a couple of seconds after the main page loads.

[kenrussell]

Thanks @kainino0x for reviewing.

I'd like to give everyone time to review this one especially given that @greggman found the earlier issue which motivated the change to an iframe.

[greggman]

lgtm, just a suggestion

[greggman]

how about

<iframe src="https://web3dsurvey.com/collector-iframe.html" width="1" height="1" style="border: none;">

IIUC, the code as is is asking for a 300x150 iframe to be scaled to 1x1 where as the suggested code is asking for a 1x1 iframe. Also frameborder="0" is deprecated.

[kainino0x]

AFAICT the width/height attributes do the same thing as the CSS, on iframes. So could just as well be all CSS.

[greggman]

You're right! Thanks!