I was interested in Angle for exactly this same reason — it would allow us to expose OpenGL at the public API level. However there are licensing issues we’d have to look at, performance tests to be run, security audits performed, and whether or not it is actually able to perform well.
Although the browsers use it for WebGL, WebGL is not the main thing browsers do. What I mean by that, is that if WebGL isn’t working, an HTML author can detect that and redirect or provide some kind of error to the user. If GL doesn’t work for us, we’d be dead in the water (probably just crash) without having some kind of fallback. We could maybe just fallback to software rendering (and realize that in such cases the performance will not be good and people will be mad). It didn’t look like a slam dunk to me. Rather, it seemed to me that we should allow the OpenGL stack to run on Windows with an option, let developers opt into it, but note that it isn’t a supported configuration so we don’t have support costs associated with it if it doesn’t work. And we’d have to forbid it on WebStart / Applets (within reason) so as not to allow bugs in the native drivers to be exploitable through us (if the board causes the VM to crash, there is potentially some security issues there). And then expose an API that works with GL, supported on Mac / Linux, but “known to work” on Windows in cases where Windows GL support works. That seemed to me a shorter path to victory. Richard On Jul 21, 2014, at 1:13 PM, Joseph Andresen <joseph.andre...@oracle.com> wrote: > That's a good point Robert, > > If the GLContext work that steve and felipe did become an actual thing, this > would help that cause become cross platform. > Angle also is strictly es2, and I haven't looked at prism es2 in a while but > I think we use GL2 calls for desktop in some cases. We would have to address > those cases (if even possible) before any work started. > > -Joe > > On 7/21/2014 10:40 AM, Robert Krüger wrote: >> On Mon, Jul 21, 2014 at 7:09 PM, Joseph Andresen >> <joseph.andre...@oracle.com> wrote: >>> I also forgot, >>> >>> The argument could be made that if we did indeed use angle, we could ditch >>> our directx 9 pipeline altogether and just use "one" hardware pipeline. We >>> would really have to evaluate this though, and I am not sure the work would >>> be worth the benefit (if there even is any). >> Well, at least the presence of the directx pipeline was used as an >> argument against exposing a GL context via a low-level native api, >> which quite a number of people with particular graphics/performance >> requirements need IIRC, so this would be a potential benefit. >
