- From: Brian Birtles via GitHub <sysbot+gh@w3.org>
- Date: Thu, 20 Sep 2018 22:43:06 +0000
- To: public-svg-issues@w3.org
I'm really not keen on 3. There is currently no concept of "when needed" and it introduces a lot of complexity. For example, is normalization needed when you have a timing function of `step-end`? Is it needed when you sample at exactly one of the keyframe offsets? Is it needed when you are filling? Is it needed when a path is combined with another value that makes the whole value fall back to 50% flip behavior? Different engines have different answers to these questions. In order to implement any answer to the above we will need to add special case code that feels to me like a layering violation (e.g. passing flags around to say, "Are we in an animation path in a particular animation state where we need to return a normalized value"). Furthermore, from an authoring point of view having `getComputedStyle` return absolute paths for some set of the above situations, but not others is surprising and seems to violate the idea of avoiding discontinuities. My experience is that adding complexity to interpolation code almost always creates more work over time (particularly as the set or permutations continues to grow) and should be done at computed value time if possible. This is particularly so as we try to extend this behavior through additive CSS, additive animations, collapsing forwards-filling animations etc. -- GitHub Notification of comment by birtles Please view or discuss this issue at https://github.com/w3c/svgwg/issues/321#issuecomment-423357803 using your GitHub account
Received on Thursday, 20 September 2018 22:43:08 UTC