SVG Working Group Teleconference

26 Jan 2012


See also: IRC log


+1.206.734.aaaa, ed, cabanier, heycam, krit, cyril, Tav, ChrisL, [Microsoft]


<trackbot> Date: 26 January 2012

<krit> Zakim: and me?

<krit> Zakim: who's here

<scribe> scribenick: cabanier

CSS Transforms and the behavior of SVG DOM

<ed> http://www.w3.org/Graphics/fx/wiki/Merged_Transforms#CSSOM_issues

krit: I'm working on the merged css transform specification

<krit> http://www.w3.org/Graphics/fx/wiki/Merged_Transforms

krit: I found some possible SVG DOM issues
... issue 11

Issue 11: Interaction of SVG DOMs SVGTransformList and SVGTransform to the CSS property

On transforming the SVG attribute transform' to a presentation attribute, the interaction of SVG DOM with CSSOM must be clarified. While an SVG attribute has a one-dimensional hierarchy where the value can just be influenced by the unique SVGTransformList, CSS had different cascading. Like described for CSSOm above, it is difficult to map cascading to the one-dimensional hierarchy, since different

style sheets can affect the style of an element and there for can have different affects to computed style. Computed itself style doesn't allow modifications.

krit: how can SVG DOM work with CSS?
... my proposal is to add a new stylesheet

<krit> Follow other presenattion attributes with transform

<krit> Creating new Author style sheet

<krit> the hierarchy of the author style sheet is after UA style sheets and before other author style sheets

<krit> SVG DOM should reflect the new created style sheet

heycam: this is just make the SVG DOM expose the stylesheet

<heycam> the style sheet for the presentation attributes (the low specificity one) that is

heycam: on first reading this sounds reasonable
... people might be confused that CSS rules are not reflected in SVG DOM
... we had a discussion that make the base value the initial value

<ed> it was "inline style" not "initial value", no?

<ChrisL> yes

heycam: reading out the value would get you the animated transform

chrisl: animval is the computed val?

<ChrisL> so the base value reflects the presentation attribute and the animval reflects the computed value

heycam: basevalue and animval can reflect different thing

krit: animval is the computed value

ed: that makes most sense but I'm concerned about whether that means a flattened matrix, or if it's a list of matrices

krit: that is not defined at the moment
... it makes most sense to have a list of animated values as well

<ChrisL> agree a tranfrom list is more useful

heycam: it is more useful to inspect to have a list of transform item rather than a flat matrix

chrisl: you can always create a matrix, but no the other way round

cyril: why is this so different?
... is it because of the cascade?

kris: yes, the transform property has the cascade

heycam: transform is now becoming a presentation attribute and we have to figure out what that means for the SVG DOM
... there is no access to the CSS dom from SVG. There is no .fill attribute

krit: there is no animval/baseval to fill

heycam: I agree with Dirk's proposal. It matches what I was thinking of.

<heycam> I prefer Dirk's suggestion because if you don't use transform in style="" or in CSS style sheets, then the SVG DOM for transform works the same as it always has.

heycam: it also works with the animval
... by having animval reflect the computed style, it will looks the same as before

cyril: does patrick's proposal to turn some attributes into properties has the same problem ? If so, is the current solution also applicable?

heycam: yes, we would want a consistent solution

ed: what does 'works with animval' mean?

<ChrisL> yes we do

<ed> issue 12

krit: can we do 3d transform on SVGTransformList?
... we would need new interfaces for perspective, etc

<ChrisL> yes, we would

krit: do we want to extend SVG interfaces to access 3d transforms?
... we would have to specify everything that is in CSS, in SVG as well

heycam: I don't think it's that much work, since it pretty simple
... otherwise we would have to specify what happens when you read out the transformlist

krit: I would specify that it's unresolvable. SVG DOM would just return undefined
... you would access everything with CSS and this would make things backward compatible

heycam: I guess we'll have similar issue with other issues
... ie by using 'calc' in css

krit: yes, it is a general issue
... I would not develop SVG DOM for these use cases

heycam: yes, maybe we'll use the CSSOM to manipulate
... if we improve the SVG DOM, should we target transform or should we do that in CSS?

krit: just do it in CSS

heycam: ie accessing width/height of a rectangle

krit: if you have SVG length, you can drop the 'px' but not in CSS
... I think the group should work on allowing that in CSS

heycam: in CSS you can using units, while in SVG you can't
... ie you can use '%' in CSS while in SVG you can't

krit: maybe we should resolve % to pixel in SVG

heycam: that would be possible
... I'm OK with returning unknown type

ed: in practice, people will use the CSSOM
... since if you use the CSS feature, you will want to use that interface anyway
... we should try to make the CSSOM as good as possible

krit: yes
... I would like to have the same features in CSSOM that we have in SVG now
... make the CCSS interface such as transformlist live

heycam: does the combined transforms spec define any CSSOM interfaces?

krit: we have not specified transformlist but have css matrix

heycam: so if you use getcomputedstyle you will get them back?

krit: changes to cssmatrix are reflected in css transform as well.
... css transforms are more complicated than other css features

heycam: this is more a question for the CSS people

chrisl: it's not clear who is editing that.

heycam: glenn is supposed to be editing that

resolution: for issue 11 and 12, we will accept the recommended solutions (see: http://www.w3.org/Graphics/fx/wiki/Merged_Transforms#SVG_DOM_issues)

OpenType and SVG

krit: the problem are not willing to implement SVG fonts

chrisl: I proposed that we implement only WOFF
... it's very clear that mozilla and microsoft are not going to implement this
... there is a proposal to add SVG fonts to OpenType
... my feeling is that is longer term than SVG2

heycam: we have an intern working on this.
... SVG glyphs inside OpenType

chrisl: there is no consensus how this format will look like

<krit> https://wiki.mozilla.org/SVGOpenTypeFonts

heycam: let me give you a link to the bug

<heycam> https://bugzilla.mozilla.org/show_bug.cgi?id=719286

heycam: I have never seen that page before

chrisl: me neither

heycam: this looks more like brainstorming

chrisl: this page should be updated with pointers to mailing lists, etc

heycam: the idea for the intern is to see how this works and report back

chrisl: I'm worried that the message will be that we already figured it out
... experimentation is great though. I'm very pleased
... sairus from Adobe is working on this

cabanier: I believe Sairus has a proposal. I can ping him.

<heycam> http://www.w3.org/community/svgopentype/

dirk: we need to find a way to have a public discussion ie by a mailing list

chrisl: I started the community group because people were emailing many groups

<heycam> http://lists.w3.org/Archives/Public/public-svgopentype/

chrisl: have a central location that is archived is important
... woff encapsulates OpenType
... CSS3 also mandates OT
... we are also going that way

<ChrisL> so if opentype adds this, svg will use it. discussion should be on the community group

resolution: any discussion should happen on the SVG OpenType group (http://www.w3.org/community/svgopentype/)

SVG2 Requirements

<ed> http://www.w3.org/Graphics/SVG/WG/wiki/SVG2_Requirements_Input

cyril: I will edit the wiki page

<cyril> http://www.w3.org/Graphics/SVG/WG/wiki/SVG2_Requirements_Input#the_.3CsolidColor.3E_element


chrisl: it's a single element gradient
... it's for paint servers, that are pointed to by a url

krit: can't this be done with CSS right now?

<ed> <solidColor id="solidMaroon" solid-color="maroon" solid-opacity="0.7"/> <rect fill="url(#solidMaroon)" .../>

chrisl: not sure. by having the same class and style rule

chris: css animation would select a class, right?

krit: yes
... if you use a selector that is less specific, does the animation not happen?
... regardless of using this in CSS, people today are working around this with gradients.

tav: inkscape has a horrible patch to do this with a one stop gradient

chrisl: a solid color is really what you want to do. SVG 1.1 had it in there. It seems like a solid feature

<ChrisL> and 1.2T had it too

tav: inkscape would have a much simpler solution if it was in there

<ChrisL> and inkscape needs it

cyril: could inkscape define a color palette with CSS classes?

<ed> so essentially it transforms <linearGradient><stop stop-color="foobar"/></linearGradient> into <solidColor solid-color and solid-opacity property="foobar"/>...

tav: inkscape doesn't do classes

<heycam> I think it you couldn't just do this with CSS, since you either need to have various different style rules for different properties, so you lose the ability to change the colour in a single place.

heycam: CSS variables seems to be solving the same use case

chrisl: having named colors

is one of the primary reason for them. This is a very useful feature.

<ChrisL> it regularises the syntax so that you can use a paint server for solid as well as radient and pattern fills.

heycam: CSS variables is still far away but it would be good to avoid the same feature twice

<ChrisL> also, you can animate the url to go from a solid to a gradient ad back

<ChrisL> its very flexible

ed: this is very simple to implement. It doesn't add much code

resolution: we will add solidColor element to SVG2


<cyril> http://www.w3.org/Graphics/SVG/WG/wiki/SVG2_Requirements_Input#xml:id


chrisl: we should drop it

everyone: agreed

resolution: drop XML:id

role, rel, rev, about, content, datatype, property, resource, typeof

chrisl: was added to support RDFa and microdate, but hasn't seen much use but it is used in HTML

heycam: doesn't exactly match with what HTML5 has since they start with item

krit: I think it's great but should match HTML5

heycam: HTML5 doesn't require RDFA and microdata
... microdata came with a set of DOM interfaces that we want to support

chrisl: yes

heycam: scanning the HTML5 spec, I don't see where RDFA is required.
... maybe there is a seperate spec

chrisl: RDFlite is I believe
... it could be that the spec hasn't been updated

resolution: for RDFa and microdata follow what HTML does


<ed> this depends on <textArea>

chrisl: this is moot since we're following CSS
... not textarea

<ChrisL> its tied to f the 'textArea' element. which we are not using

resolution: we won't display-align since we won't have textarea

buffered rendering


ed: it's a hint to cache as much as possible
... it's not a requirement

cabanier: this is very useful

krit: I'm fine with it

heycam: I'm concerned about these hints
... like filter-region

ed: I agree

<ed> s/agree/share your concern about misuse of this property, but it's sometimes needed for specific devices/scenarios/

resolution: after implementor feedback, we agree that buffer rendering hints are needed

heycam: with CSS transform, you can get pixelated buffers
... would we have a requirement that they should rerender

ed: in most cases you don't want pixels

Summary of Action Items

[End of minutes]

Minutes formatted by David Booth's scribe.perl version 1.136 (CVS log)
$Date: 2012/01/26 21:38:29 $

Scribe.perl diagnostic output

[Delete this section before finalizing the minutes.]
This is scribe.perl Revision: 1.136  of Date: 2011/05/12 12:01:43  
Check for newer version at http://dev.w3.org/cvsweb/~checkout~/2002/scribe/

Guessing input format: RRSAgent_Text_Format (score 1.00)

Succeeded: s/priority/specicicity/
Succeeded: s/specicicity/specificity/
Succeeded: s/baseline/base value/
Succeeded: s/that makes most sense/that makes most sense but I'm concerned about whether that means a flattened matrix, or if it's a list of matrices/
Succeeded: s/patrick's proposal has the same proplem/does patrick's proposal to turn some attributes into properties has the same problem ? If so, is the current solution also applicable?/
Succeeded: s/on SVG as well/on SVGTransformList/
Succeeded: s/stop developing/not develop/
Succeeded: s/do it will/define a color palette with/
Succeeded: s/wouldn't work with CSS/you couldn't just do this with CSS/
Succeeded: s/solid-color/solidColor/
Succeeded: s/solid-color/solid-color and solid-opacity property/
Succeeded: s/hasn't seen much use/was added to support RDFa and microdate, but hasn't seen much use/
Succeeded: s/scannig/scanning/
WARNING: Bad s/// command: s/agree/share your concern about misuse of this property, but it's sometimes needed for specific devices/scenarios/
Succeeded: s/we should/would we have/
Succeeded: s/have have/have/
Succeeded: s/get/want/
Found ScribeNick: cabanier
Inferring Scribes: cabanier
Default Present: +1.206.734.aaaa, ed, cabanier, heycam, krit, cyril, Tav, ChrisL, [Microsoft]
Present: +1.206.734.aaaa ed cabanier heycam krit cyril Tav ChrisL [Microsoft]
Agenda: http://lists.w3.org/Archives/Public/public-svg-wg/2012JanMar/0038.html
Found Date: 26 Jan 2012
Guessing minutes URL: http://www.w3.org/2012/01/26-svg-minutes.html
People with action items: 

WARNING: Input appears to use implicit continuation lines.
You may need the "-implicitContinuations" option.

[End of scribe.perl diagnostic output]