W3C home > Mailing lists > Public > public-fx@w3.org > January to March 2012

Re: Adding normal matrix to CSS Shaders

From: Fabrice Robinet <cmg473@motorola.com>
Date: Thu, 12 Jan 2012 10:45:33 -0800
Cc: "Gregg Tavares (wrk)" <gman@google.com>
Message-Id: <E3FEEA20-4279-4E19-BAEB-4CA77DDF9254@motorola.com>
To: public-fx@w3.org
...mail went away to quickly...  full answer rephrased.. sorry for the noise

+ in addition to my previous comment.

That was probably misleading from me, to get GL as reference.
While GL tends to become more general and get rid of the rest of the fixed pipeline and some associated built-in uniform (like normal matrix).
An opposite move with spec like CSS Shaders is to be more specific and rely on clear / meaningful attributes. 

-> That's why I believe that normal matrix could fit well eventually.
It's also okay to have alternatively more general matrixes attributes like "inverse transpose", 
but as I stated in the mail before , that's many choices to take if we don't want take all the combinations. 
So what I didn't like with the totally generic matrix description (such as NVIDA)  if that I felt it a bit premature (wrt to the SPEC)  and heavy to sort it out well.


> 
> On Jan 12, 2012, at 10:26 AM, Fabrice Robinet wrote:
> 
>> Hi Gregg,
>> 
>> inline,
>> 
>> On Jan 12, 2012, at 10:09 AM, Gregg Tavares (wrk) wrote:
>> 
>>> 
>>> 
>>> On Tue, Jan 10, 2012 at 10:46 PM, Fabrice Robinet <cmg473@motorola.com> wrote:
>>> Hi all,
>>> 
>>> Let's revive this thread after christmas break ;)
>>> 
>>> Back to the original proposal around the normal matrix,
>>> Indeed,  "normal matrix" is less "general" than "inverse transpose". 
>>> But I think "normal matrix"  it is preferable as it allow to stay conservative (its addition can be done independently of anything else) 
>>> as opposed to "inverse transpose" which would involve to take premature decisions about other combinations that one "might need".
>>> - Normal matrix typical use-case is for lighting, and that's clear. -
>>> 
>>> Moreover, because of the similar (old) built-in gl_NormalMatrix uniform in GLSL,
>>> I believe the "normal matrix" would seem familiar/natural for most shaders aficionados.
>>> 
>>> I disagree. GL has been trying to get away from the fixed function specialized cruft of the last 18 years. That's why gl_NormalMatrix is not in OpenGL ES at all and has been deprecated in current OpenGL 
>> 
>> Yes, but how GL is implemented being totally programmable like ES2 or not partly fixed as GL on desktop nowadays is not really my point here.
>> 
>> What I was saying is that the - concept - or normal matrix is something that shader developers are used too.
>> 
>> Providing a new supported uniform in the context of CSS Shaders wether it comes from ES2 or not, does not really matters.
>> 
>> These uniform whose name carries semantic as well in the CSS Shaders are something that looked to be agreed when building the CSS Shader SPEC,
>> so the normal matrix would just follow the same pattern.
>> 
>> This new built-in as just to be consistent with the rest of the CSS Shaders SPEC, and would be named in that case somehting like "u_normalMatrix"
>> 
>>> Let's not go adding that cruft back in.
>>> 
>>>  
>>> 
>>> 
>>> Let's continue the discussion if there are more concerns about this,
>>> otherwise that would be awesome to have this uniform supported for a new revision of the spec.
>>> 
>>> As a reminder, there is a bug tracking this:
>>> https://www.w3.org/Bugs/Public/show_bug.cgi?id=15253
>>> 
>>> Thanks,
>>> Fabrice.
>>> 
>> 
> 
Received on Friday, 13 January 2012 19:21:45 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Friday, 13 January 2012 19:21:46 GMT