After some discussion, the GFX group thinks it has decided what it definitely wants to get accomplished during Q2 of 2010. This includes, but is not necessarily limited to:
* Improve interactive performance by shipping, in a developer preview: ** Layers implementations based on OpenGL and OpenGL ES; and ** Direct2D support, turned on for at least some subset of hardware. (This depends on RelEng - {{bug|549120}}!) * Improve CSS3 font support by implementing Harfbuzz for simple scripts on Tier-1 platforms in a developer preview, and creating a CSS3 font test suite.
By "shipping in a developer preview," we mean something that is alpha-quality; that is, everything might not work yet, but it won't stop someone from using a Minefield Developer Preview alpha build. This also implies that all of our automated testing passes on a build with such a feature enabled, on a machine with support for that feature.
Our stretch goal internally is to focus on OpenGL ES (for use on mobile) such that it's ready for general consumption much earlier than the other layers implementations. This, and potentially Direct2D, are what the hardware acceleration side of the graphics group will be focusing on.
My current understanding is that we will explicitly be de-emphasizing work on the Direct3D 10 layers implementations during this quarter, both because the same resources will be needed for Direct2D work, and most interactivity improvements will be seen by users whose computers are already capable of Direct2D.
Harfbuzz shaping is going to take a while, which is why our font folk John and Jonathan are focusing only on getting simple script up and running this quarter. We also want a CSS3 font test suite in order to push the CSS3 font spec into standardization.
On Tue, Mar 30, 2010 at 11:24 AM, Joe Drew <j...@mozilla.com> wrote:
> My current understanding is that we will explicitly be de-emphasizing work > on the Direct3D 10 layers implementations during this quarter, both because > the same resources will be needed for Direct2D work, and most interactivity > improvements will be seen by users whose computers are already capable of > Direct2D.
For those unaware, Wan's talking about Direct3D 9 layers.
Wan, honestly, I don't know. We might prioritize it lower because it's older hardware. We might prioritize it higher because people with Direct3D 9 won't get the gains of Direct2D. It didn't even come up in our discussions, so for now, I don't think the GFX group will be working on it -- but patches are always accepted!
On Tue, Mar 30, 2010 at 11:24 AM, Joe Drew < j...@mozilla.com > wrote:
My current understanding is that we will explicitly be de-emphasizing work on the Direct3D 10 layers implementations during this quarter, both because the same resources will be needed for Direct2D work, and most interactivity improvements will be seen by users whose computers are already capable of Direct2D.
> By "shipping in a developer preview," we mean something that is > alpha-quality; that is, everything might not work yet, but it won't stop > someone from using a Minefield Developer Preview alpha build. This also > implies that all of our automated testing passes on a build with such a > feature enabled, on a machine with support for that feature.
I'm very worried about this part. The current release roadmap, as I understand it, is to do a beta of 1.9.3 in mid-June with a scheduled release in October. By targeting an alpha for these goals you're implicitly missing this release vehicle. Is that intentional?
I think that the OpenGL layers, partly targeted for Maemo/Fennec, need to be beta-quality by the end of the quarter to meet our Mobile ship schedule.
I think that many of the other features could possible be shipped on subsequent stable branch updates (1.9.3.x), but I think we need to be clearer about how these goal statements interact with the ship schedule.
> > By "shipping in a developer preview," we mean something that is > > alpha-quality; that is, everything might not work yet, but it won't > stop > > someone from using a Minefield Developer Preview alpha build. This > also > > implies that all of our automated testing passes on a build with > such a > > feature enabled, on a machine with support for that feature.
> I'm very worried about this part. The current release roadmap, as I > understand it, is to do a beta of 1.9.3 in mid-June with a scheduled > release > in October. By targeting an alpha for these goals you're implicitly > missing > this release vehicle. Is that intentional?
Obviously we'd rather everything be beta quality by then, but we just don't think we can commit to it with the resources we have. We're going to try our best, and we might get lucky, but otherwise, it'll have to be in a point update to 1.9.3.
> I think that the OpenGL layers, partly targeted for Maemo/Fennec, need > to be > beta-quality by the end of the quarter to meet our Mobile ship > schedule.
That's why we're going to focus on OpenGL ES at the exclusion of D3D10. I don't know how realistic getting it to beta by the end of the quarter is, but it's what we're aiming for internally.
