[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Public WebGL] experimental and prefixes still prevalent



If you don't mind clarifying: What do you mean by "shown with the experimental flag?" Does this mean:

 - The browser will accept "experimental-webgl" as a valid context type? (I think all WebGL enabled browser match this description)
 - The browser ONLY accepts "experimental-webgl"? (This sound like what you're talking about)
 - WebGL support in any form must be enabled with a flag on those browsers?

In the case of prefixed extensions, Chrome's policy is to no longer launch extensions with prefixes, though we will keep the prefixed versions around for older extensions. That doesn't necessarily apply to other browsers, however. I think Firefox said they would do the same as Chrome, but in Apple's case if they don't want to introduce a flag mechanism into the browser (which I'm sure they don't on iOS) prefixed extensions may be the only way they can expose draft extensions. Not that either of the cases you mentioned are still in draft status, I'm just saying there's still valid reasons to launch with a prefix.

I think in the case of ATC textures we had implemented the extension back when prefixing was a thing, but there may have been some errors in the implementation that prevented it from showing up in the stats till more recently.

--Brandon

On Fri Oct 10 2014 at 1:15:00 AM Florian Bösch <pyalot@gmail.com> wrote:
WebGL is still shown with the experimental flag on some platforms, where it probably shouldn't. To illustrate the issue, below are some of the select cases:

  • Opera is entirely experimental, though it is based on Webkit (or Blink?) which is in use by Chrome, which generally has a very low experimental webgl status.
  • Firefox has started to go out of experimental, but the rate of change has slowed and seems to be flat now, this would indicate that browser update isn't working satisfactory.
  • Safari on OSX is experimental 30% of the time, though it is based on WebKit, as is Chrome which for the same platform is almost never experimental.
  • Chrome on Linux shows 6% as experimental, although Chromium on Linux only shows about 2% as experimental (this seems strange)
  • Android Browser on phones is 61% experimental, but Chrome Mobile is 0% and the Android Browser on tablets is 
  • Android Browser on phones is 61% experimental (Chrome Mobile is 0% experimental, Android browser on tablets is 6%)
  • Mobile Firefox on Android is almost completely experimental (but has recently started to go out of experimental)
There are also still some extensions being listed with vendor prefixes (but are declining), but more importantly, some extensions seem to start out their life again with vendor prefixes, and I thought the policy was not to do that anymore. Below are cases where this happens:
  • WEBGL_compressed_texture_atc:  with prefix on Android by Chrome and Android browser.
  • WEBGL_compressed_texture_pvrtc: with prefix on iOS 
  • EXT_texture_filter_anisotropic: with prefix on iOS.
You can inspect the current state of these on http://webglstats.com/ for each platform, operating system and browser using the selection tree on the side. I'd encourage all vendors to have a look at situations like this, and see if it can be improved.