Re: Proposed wording for 'how object transforms affect gradient rendering output'

Diagrams to go with the email.

Anthony Grasso wrote:
> Hi SVG WG,
> 
> Firstly, thank you to Cyril for the valuable feedback provided. Please 
> see my comments below.
> 
> Cyril Concolato wrote:
>> Hi Anthony,
>>
>> Here are some comments.
>>
>> In your wording you use the term 'initially'. I think I understand 
>> what it means but I would prefer if you would say something like: 
>> "before any transformation is applied to the gradient".
>>
> 
> Agreed. The suggested wording has been used in the new draft of the 
> proposed wording (end of email).
> 
>> You also mention the magnitude of the normal. I don't understand what 
>> it means and how it is used. To me, it is sufficient to say that the 
>> normal is perpendicular, we don't even need to say the direction +90° 
>> or -90°. It does not matter.
>>
> 
> True. The part about the magnitude of the normal has been removed.
> 
>> Concerning the order between object transforms and gradient 
>> transforms, are you sure about it? I thought it was first apply 
>> gradient transforms because they apply to all reference of the 
>> gradients, then object transforms. If you look at SVG 1.1 [1], you'll 
>> find:
>> "[The gradientTransform] transformation matrix is post-multiplied to 
>> (i.e., inserted to the right of) any previously defined 
>> transformations, including the implicit transformation necessary to 
>> convert from object bounding box units to user space."
>>
> 
> Right again. The wording has been adjusted accordingly to match what it 
> is currently written in the standard.
> 
>> Concerning the position of the text. I think the first two sentences 
>> "The angle ... vector." only apply to linear gradients because of the 
>> notion of gradient vector. For the rest of the paragraph, it's fine to 
>> put it in the introduction.
>>
> 
> A gradient normal is useful when calculating skews on radial gradients 
> as it provides the tangent at each colour point on the gradient. Please 
> see the attached diagrams.
> 
> The revised wording of the first paragraph in the proposed text now reads:
> 
> [[
> The angle of the color transitions along the gradient vector is defined 
> by the gradient normal. Before any transforms are applied to the 
> gradient or its referencing objects, the gradient normal is 
> perpendicular with the gradient vector. If an object references a 
> gradient, conceptually the object should take a copy of the gradient 
> vector and gradient normal and treat it as part of its own geometry. Any 
> transformations applied to the object geometry also apply to the copied 
> gradient vector and gradient normal. Any gradient transforms that are 
> specified on the reference gradient are applied before any object 
> transformations are applied to the gradient.
> ]]
> 
> 
> Please advise before the next telcon if there is any more wording that 
> needs to be adjusted or if  any other changes need to be made.
> 
> Erik, can we add this issue to the next telcon agenda please.
> 
> 
> Thanks,
> 
> Anthony
> 
>> Regards,
>>
>> Cyril
>>
>> [1] 
>> http://www.w3.org/TR/SVG11/pservers.html#LinearGradientTransformAttribute
>>
>> Le 04/02/2010 05:58, Anthony Grasso a écrit :
>>> Hi SVG WG,
>>>
>>> In relation to the discussion on "gradient userSpaceOnUse" [1] and the
>>> investigation I did on "how authoring tools apply object transforms to
>>> gradients" [2], I have devised some wording to put in (possibly) current
>>> and future SVG specifications.
>>>
>>> I propose adding the following wording and diagrams for the different
>>> parts in the gradient section:
>>>
>>> Wording to be placed after the first paragraph of the introduction for
>>> gradients [3] (in SVG Tiny 1.2)
>>>
>>> [[
>>> The angle of the color transitions along the gradient vector is defined
>>> by the gradient normal (see diagram below). Initially, the gradient
>>> normal is perpendicular with the gradient vector and is the same
>>> magnitude as the gradient vector. If an object references a gradient,
>>> conceptually the object should take a copy of the gradient vector and
>>> gradient normal and treat it as part of its own geometry. Any
>>> transformations applied to the object geometry also apply to the copied
>>> gradient vector and gradient normal. Any gradient transforms that are
>>> specified on the reference gradient are applied after object
>>> transformations have been applied but prior to rendering.
>>> ]]
>>>
>>>
>>> The wording in the last sentence of the second last paragraph for
>>> gradientUnits attribute [4] should be reworded to:
>>>
>>> [[
>>> When the object's bounding box is not square, the gradient normal which
>>> is initially perpendicular to the gradient vector within object bounding
>>> box space shall render non-perpendicular relative to the gradient vector
>>> in user space. This transformation is due to application of the
>>> non-uniform scaling transformation from bounding box space to user 
>>> space.
>>> ]]
>>>
>>>
>>>
>>> Note that this wording is a first draft and open to discussion. To my
>>> understand of the issue, there are some implementations that have an
>>> alternative method of applying transforms and thus producing a different
>>> result. I can sympathise that this wording is potentially contentious,
>>> but given that gradients are an integral part of any graphics language,
>>> a consensus must be reached on this issue so all implementations will
>>> produce the same result.
>>>
>>> Cheers,
>>>
>>> Anthony
>>>
>>>
>>> [1] http://lists.w3.org/Archives/Public/www-svg/2010Jan/0000.html
>>> [2] 
>>> http://lists.w3.org/Archives/Public/public-svg-wg/2010JanMar/0035.html
>>> [3] http://www.w3.org/TR/SVGTiny12/painting.html#Gradients
>>> [4]
>>> http://www.w3.org/TR/SVGTiny12/painting.html#LinearGradientUnitsAttribute 
>>>
>>
>>
>