From: Øyvind Stenhaug <oyvinds@opera.com>
Date: Wed, 29 Feb 2012 17:58:06 +0100
To: "Boris Zbarsky" <bzbarsky@mit.edu>, "Tab Atkins Jr." <jackalmage@gmail.com>

Message-ID: <op.waf9e4i6bunlto@oyvinds-desktop>
```On Tue, 28 Feb 2012 17:16:28 +0100, Tab Atkins Jr. <jackalmage@gmail.com>
wrote:

> On Tue, Feb 28, 2012 at 8:07 AM, Tab Atkins Jr. <jackalmage@gmail.com>
> wrote:
>> On Tue, Feb 28, 2012 at 5:12 AM, Boris Zbarsky <bzbarsky@mit.edu> wrote:
>>> On 2/28/12 3:01 AM, Kang-Hao (Kenny) Lu wrote:
>>>>
>>>> I am not sure this is well-defined as I don't know if concentric
>>>> ellipses are by definition necessarily having the same eccentricity.
>>>
>>> They're not.
>>>
>>>
>>>> I don't mean to use www-style as a Math forum but oddly enough I
>>>> couldn't find the definition of "concentric ellipses" on Google.
>>>
>>> When people say "concentric ellipses" they seem to mean
>>> non-intersecting
>>> ellipses with shared center and major axes pointing in the same
>>> direction.
>>>  Or something.  There is no commonly accepted definition I know of,
>>> other
>>> than "I know it when I see it".
>>
>> Huh.  I thought there was a simple definition (linearly scale both
>> axises).  I'll do some quick testing to see what impls do and ensure
>> it matches what I assume, then specify that more clearly.
>
> Yup, everyone does the obvious thing.  I've added the word
> "proportional" in front of "concentric ellipses".  That sound good?

I think you'd need a further restriction. As I understand it, a radial
gradient consists of all ellipses E such that
1) E and the ending shape are concentric - that is, they have the same
center
2) E and the ending shape have the same eccentricity - I assume this is
what "proportional" means, though using that word when talking about
two-dimensional shapes sounds unfamiliar to me (I didn't use the term
"similar" since that has a non-mathematical meaning which is much more
vague)
3) the major semiaxes/radii of E and the ending shape coincide (otherwise
you'd include ellipses that are rotated by any arbitrary amount)

(I also think that the orientation of the ending shape's axes is only