steve at sjbaker.org
Sat Dec 18 20:17:54 PST 2010
On 12/18/2010 12:58 PM, Zack Morris wrote:
> On Dec 18, 2010, at 10:53 AM, Martijn Buijs wrote:
>> I hope this web implementation will be based on OpenAL-soft and thus entirely software rendered on the PC platform. From my experience, the quality of Creative's hardware implementations of OpenAL have remained consistently disappointing. At least use OpenAL-soft as reference implementation.
It's far too early to be coming down on one side or the other here.
Right now, this is no more than my personal wish...step one is to get
the stake-holders around a virtual table to discuss it.
> I think what is really needed is a unit test of some kind, where if someone wants to write a hardware renderer, the output should be recorded and compared against the software implementation and if the difference is more than say 5% or something, the hardware renderer couldn't be certified as actually being OpenAL.
If the WebGL effort is anything to go by - having a comprehensive test
and compliance suite is a huge part of the work. So, yeah - that would
certainly happen if we could make this idea stick.
> I spend most of my time now just testing the simplest concepts like FBOs and shaders in OpenGL (which are really just registers and transformations, if you look at them from an engineering standpoint) and I am constantly amazed how often specifications just completely fall down. If a specification allows falling back to software, then a game can't use that case at all, because poor performance is a failure itself.
Yeah - that's a massive issue...and a lot of it is politics. When you
get a company like Intel making graphics chips that are truly *nasty*
devices - there ends up being a ton of money involved. If Intel can't
make their chips pass the basic Windows-7 minimal requirements - then
they won't sell a single chip. On the other hand the makers of
motherboards, low end laptops and netbooks don't give a damn about
quality - they just want a machine that has the Microsoft seal of
approval at lowest price. Those pressures force lax specifications -
and if you tighten up the spec then the vendor is forced into horrible
What you really want is savvy customers who don't buy crap...but that
doesn't seem to be stopping Intel.
> I would really like it if the world got back to hard specs like "this is how VGA works" or "this is how to make a USB device be HID compliant". I'm finding more and more often that that level of reliability only comes from open source. It's a pity too because Nvidia, ATI, Apple and Microsoft are the ones with the billions and billions of dollars. They should do better.
Perhaps - but then it is necessary for the spec-writers to design the
technology - and in these days of GPU's and such - that's a massive
Anyway - I'm fairly pleased with the way that WebGL has turned out
(well...is continuing to turn out...it's not finished yet) - if you have
a moment, install the latest nightly build of Firefox or Chrome and head
over to Google's new "BodyBrowser" site....truly stunning stuff.
More information about the Openal-devel