SVG WG Sydney 2012 F2F day 3 -- 12 Jan 2012

<cabanier> scribenick: cabanier

alpha mask proposal

<heycam> http://www.w3.org/Graphics/SVG/WG/wiki/Proposals/Alpha_mask

cm: mask are currently unintuive since they use luminance
... it is quite confusing
... we opened an issue where you can use the alpha values
... I opened a small proposal where you can specify luminance or alpha
... there were a proposal where you could combine the 2 by having a seperate property for each mask type

rc: are alpha values currently ignored?

cm: no they contribute to the luminance.

rc: this is similar to what PDF
... except noone uses the alpha version

bb: there seems to be some confusion in the spec about what mask apply to

it seems pretty unlikely for a given mask mask element that it can be use for both luminance and alpha

rc: why is that?

cm: it's unlikely that you use the same element but then use it with alpha in one place and luminance in the other

<birtles> (re the confusion in the spec, specifically http://www.w3.org/TR/SVG11/masking.html#MaskProperty says the mask property can refer to "another graphical object", not just a <mask>. That restriction appears earlier in the prose.)

<ed> another way could be mask: <funciri> | none | alpha(<funciri>) | luminance(<funciri>)

<ed> replace those <funciri> in the last two with <iri> I guess though

<ed> that would make it rather easy to do fallbacks because old useragents would fail parsing the css

bb: it would be nice if you wouldn't have to create id's

cm: not everyone puts masks in the defs
... that does seem handy

bb: very often you have to write a unique id generator to create the id's

cm: maybe we can raise an issue

ed: people have asked for this for gradients

<heycam> ISSUE: Consider allowing <mask>, <clipPath>, etc. as direct children of elements they'd apply to, to avoid IDs

<trackbot> Created ISSUE-2431 - Consider allowing <mask>, <clipPath>, etc. as direct children of elements they'd apply to, to avoid IDs ; please complete additional details at http://www.w3.org/Graphics/SVG/WG/track/issues/2431/edit .

bb: this does impace how it impacts alpha and luminance

cm: you can have new keywords for mask: 'mask: child' and then if there is a mask element child that that is using it as a child.

ed: having a keyword like that because we don't always traverse into an element. this keyword would tell us to examine the element.
... I have no problems with the functionality itself
... I think that's what we want

cm: bb do you want to investigate this further?

bb: yes, I would like to look into this further to make sure there are no issues

action birtles to investigate issue 2431

<trackbot> Created ACTION-3205 - Investigate issue 2431 [on Brian Birtles - due 2012-01-19].

action heycam to proposal a final syntax on alpha mask after Brian investigate 2431

<trackbot> Created ACTION-3206 - Proposal a final syntax on alpha mask after Brian investigate 2431 [on Cameron McCormack - due 2012-01-19].

resolution: we will add alpha mask functionality to SVG 2

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

ed: until the break we will work on some open actions

<ed> http://www.w3.org/Graphics/SVG/WG/track/users

<heycam> -- doing actions until morning tea break --

<ed> -- <tbreak dur="15min"/> --

<heycam> scribenick: heycam

SVG in HTML <head>

<ed> previous discussion of this: http://www.w3.org/2012/01/05-svg-minutes.html#item04

DS: there's strong resistance from browser vendors to add elements to the HTML <head>

… because it breaks backwards compat, and requires changes to the parser

… the parser is pretty much set in stone at this point

CL: I don't think we need to put things in head

… our <defs> is like a <div> with display:none

… so you can stick whatever you want there

… there's no need to put it in the <head>

… the only thing we do have that maybe belongs in the head is the <meta> element, though our <metadata> is structured

… in SVG you can nest elements in <metadata>

… so it really doesn't belong in <head> either

DS: I do think it's not unreasonable for people to think they should add things to the head

CL: at first sight it's very reasonable

DS: I think our best course of action is not to follow up on hixie's recommendation to define our elements as metadata content

… rather to simply to educate people that you can't do that

… I think it's very unusual that hixie would recommend this course of action

… I will respond to the original commenter

RESOLUTION: SVG will not allow elements to be put in HTML <head>

shorthand properties as presentation attributes

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

<birtles> ScribeNick: birtles

CL: CM's proposal on the wiki where you put the shorthands before all the properties in doc order satisifies my earlier concerns

CM: proposal is to allow pres attr for properties which are shorthands
... at the moment you can't write marker="" b/c marker is a shorthand
... creates confusion for authors
... don't understand why they need to use marker-start etc. if they're using pres attr
... with style you can just use "marker"
... last time this issue came up was in context of when CSS properties get split and orig becomes a shorthand
... e.g. overflow becoming overflow-x and overflow-y
... overflow is now a shorthand

CL: likewise whitespace might go that way

CM: we want to allow overflow to continue to be a pres attr
... even if it becomes a shorthand
... original reluctance was you have multiple attr targetting the same property
... and override behaviour was not clear since attr are unordered
... unlike styles where you have doc order
... but the proposal addresses that by taking shorthands first

ED: so you if have marker-mid before marker (as attributes) then marker will override marker
... so you if have marker-mid before marker (as attributes) then marker will override marker-mid

CL: it's something we're going to have to address with properties being split up and becoming shorthands
... any comments from implementors?

ED: it's probably fine but I need to check the code
... it would be pretty simple to verify if there were some test cases

CM: I think it should be ok
... is there a difference between animating the style or the presentation attribute

CL: attributeType is just for disambiguation when the attribute name and property name overlap and are not the same thing
... not to provide different behaviour

RESOLUTION: We will allow shorthand properties to be used as presentation attributes (ISSUE-2430) according to Cameron's proposal

<scribe> ACTION: Cameron to describe using shorthand properties as presentation attributes in spec [recorded in http://www.w3.org/2012/01/12-svg-minutes.html#action01]

<trackbot> Created ACTION-3207 - Describe using shorthand properties as presentation attributes in spec [on Cameron McCormack - due 2012-01-20].

ED: is there an order normally to style

<heycam> ACTION-3207: http://www.w3.org/Graphics/SVG/WG/wiki/Proposals/Shorthand_presentation_attribute

<trackbot> ACTION-3207 Describe using shorthand properties as presentation attributes in spec notes added

CL: yes, if it's in a stylesheet (incl style attr)

ED: if you start modifying the element by script, then the order is not guaranteed

CL: the assertion was attributes are unordered

CM: the stylesheet is not exposed to script though right?
... you can getComputedStyle but there's no order requirement there

ED: if it's just for when it's parsed then it's ok

CM: even if you're modifying the attr then you still need the shorthand first
... if you're actually creating stylesheets you could create two (one for shorthands)
... might be simpler than managing the order

ED: what about font? you have to use CSS rules such as requiring units
... isn't that confusing?
... for all shorthands you have to use units
... whereas elsewhere you don't

CL: it's only really for lengths where we have this different
... where you don't need units in SVG
... marker etc. are ok
... it's only really 'font'
... so it would be a rule that shorthands have to use CSS rules which includes using units (and excluding scientific notation)

CM: what if we came up with our own shorthand
... that wasn't ambiguous

CL: that would be confusing to use a different notation
... and might outweigh the benefit of using shorthands
... because you'd add syntax to disambiguate

CM: so we'll just require CSS parsing for shorthand presentation attributes

ED: I think that would be ok, better than adding quirks

CM: what about new CSS properties?
... we have font-size in SVG so we can explicitly allow unitless lengths

CL: do we want to revisit scientific notation in SVG 2?
... because perhaps it's not used?

CM: many scripts default to using exponential notation

CL: we were originally expecting much more use of the world coord system with very large numbers
... but in practice you actually run into precision problems

SVG2 Requirements

Up to: http://www.w3.org/Graphics/SVG/WG/wiki/SVG2_Requirements_Input#Improve_to-animations

<cyril> we should also deal with http://www.w3.org/Graphics/SVG/WG/wiki/SVG2_Requirements_Input#SMIL_time_containers

deciding on time containers depends on me finishing action-3202

BB: not sure what this means

CM: I guess it's about addressing differences

BB: what are the differences?

CM: underlying value is fixed, reversing

BB: my guess is reversing?

VH: we should find out what is specificallly required here

<scribe> ACTION: Cameron to contact Dr. Olaf Hoffmann to ask what is specifically required to make to-animation similar to CSS transitions [recorded in http://www.w3.org/2012/01/12-svg-minutes.html#action02]

<trackbot> Created ACTION-3208 - Contact Dr. Olaf Hoffmann to ask what is specifically required to make to-animation similar to CSS transitions [on Cameron McCormack - due 2012-01-20].

ACTION-2308: http://www.w3.org/Graphics/SVG/WG/wiki/SVG2_Requirements_Input#Improve_to-animations

<trackbot> ACTION-2308 Draft an email response for the second half of ISSUE-2089 notes added

ACTION-3208: http://www.w3.org/Graphics/SVG/WG/wiki/SVG2_Requirements_Input#Improve_to-animations

<trackbot> ACTION-3208 Contact Dr. Olaf Hoffmann to ask what is specifically required to make to-animation similar to CSS transitions notes added

CL: we are going to have transitions anyway based on Patrick's proposal to allow using SVG's attributes in CSS

<cyril> RESOLUTION: SVG 2 will allow CSS-transitions-like animations

<cyril> RRSAgent pointer

<heycam> ACTION-3208: Say to Olaf we have interpreted his request in this way, and we agree, but if he means something else please explain

<trackbot> ACTION-3208 Contact Dr. Olaf Hoffmann to ask what is specifically required to make to-animation similar to CSS transitions notes added

next, Transitions for discretely animated properties, text-anchor in particular

<ChrisL> commenter wants to smoothly animate from

<ChrisL> .................|..................

<ChrisL> SOME_TEXT_1

<ChrisL> SOME_OTHER_TEXT_2

<ChrisL> to

<ChrisL> ................|..................

<ChrisL> SOME_TEXT_1

<ChrisL> SOME_OTHER_TEXT_2

ED: even though text-anchor-end animation is discrete it would be possible to provide linear interpolation

CM: it is a bit of a violation of the model and the idea of "discrete"

<ChrisL> I think that breaks the modelof animation. at all points in the animation, the value of the animated property has to be one of the actual values

VH: there's a difference between calcMode="discrete" and calcMode="linear" that falls back to discrete

CM: if we wanted it to work for text-anchor we could allow it to take numeric values

<ChrisL> we could imaging a mapping with for example-100% as start, 0 as middle,100% as end

CM: e.g. -100 to 100

CL: we could spec it fairly easily
... I think it's ok if we're just talking about calcMode="linear" that falls back
... as opposed to calcMode="discrete"

CM: but I often rely on that fallback behaviour
... and don't specify calcMode="discrete"

CL: if we did this change and people are relying on the discrete behaviour they might get a surprise

<ChrisL> butthey could fix it by adding calcmode discrete to keep the old behaviour

BB: so you have to modify the definition of those attributes to allow numeric values

CL: the approach is to make attributes that take discrete values to be shorthand for those that take numeric values

CM: in font-weight, some of the keywords are synonyms for numbers

CL: but it's an ordered series

CM: what happens when you getComputedStyle---which do you get? the keyword or the number?
... likewise what will happen here if we use getComputedStyle on the new text-anchor etc. that takes numeric values

CL: I'm proposing promoting some values so that can take *continuous* values
... so font-weight would not be an example

RESOLUTION: SVG2 will allow linear interpolation for properties which were previously discrete

<scribe> ACTION: Chris to write up a proposal for allowing linear interpolation of properties that were previously discrete but could be reasonably interpolated linearly [recorded in http://www.w3.org/2012/01/12-svg-minutes.html#action03]

<trackbot> Created ACTION-3209 - Write up a proposal for allowing linear interpolation of properties that were previously discrete but could be reasonably interpolated linearly [on Chris Lilley - due 2012-01-20].

<scribe> ACTION: Brian to look for text in SVG/SMIL that suggests using different calcMode's based on the values used (enum values versus numeric) [recorded in http://www.w3.org/2012/01/12-svg-minutes.html#action04]

<trackbot> Created ACTION-3210 - Look for text in SVG/SMIL that suggests using different calcMode's based on the values used (enum values versus numeric) [on Brian Birtles - due 2012-01-20].

next, animateTransform with matrix-decomposed like CSS

<ed> http://www.w3.org/Graphics/SVG/WG/wiki/SVG2_Requirements_Input#animateTransform_with_matrix-decomposed_like_CSS

CM: diff between animateTransform and CSS Animations with transform property, is CSS considers the whole transform list, but SVG Animations does one transform at a time and post-multiplies
... request is to add a new type to animateTransform that matches CSS Animations
... but we could just <animate> to get that behaviour
... <animate attributeName="transform" values="scale(...) rotate(...); scale(...) rotate(...)"/>
... it overrides the whole list
... need to define how additive works with whole lists of items

VH: so how does animateTransform add to the end of the list?

BB: it differs between browsers
... if you inspect the animVal in the DOM some browsers will just append a matrix transform, some will try to preseve the animateTransform's type, others just replace the whole list with a matrix

VH: even if we deal with lists we already have a model for adding that to the underlying transform, it doesn't change the model

CC: it doesn't change the post-multiplication
... it doesn't change the final value
... but it does change the animVal

VH: but given the way the requirement is worded I think it wouldn't change
... we want to allow <animate attributeName="transform">

CM: I agree

RESOLUTION: We will support animation using a transform-list

http://dev.w3.org/csswg/css3-2d-transforms/#animation

http://brian.sol1.net/svg/animatetransform-issues/accumulation-of-transformations/

BB: Addition of transforms is a bit unusual (see above link)
... we can rely on CSS Transform's behviour for interpolation but we need to specify how addition works for transform lists

CM: this difference is specific to animateTransform
... so if we go to using animate, do we want to just use the default behaviour?
... which would mean adding the transform components rather than post-multiplication

<scribe> ACTION: Cyril to investigate different methods for adding transform in animation [recorded in http://www.w3.org/2012/01/12-svg-minutes.html#action05]

<trackbot> Created ACTION-3211 - Investigate different methods for adding transform in animation [on Cyril Concolato - due 2012-01-20].

ACTION-3211: http://brian.sol1.net/svg/animatetransform-issues/accumulation-of-transformations/

<trackbot> ACTION-3211 Investigate different methods for adding transform in animation notes added

---break 1hr for lunch---

<vhardy> http://dev.w3.org/SVG/profiles/1.2T/test/htmlObjectHarness/animate-elem-81-t.html

<patrickdengler_> Sorry, I didn't catch the end of this discussion, and I hope you had a good lunch :)

<patrickdengler_> When we talk about adding a new element, such as animateTransform, are we consulting with the HTML group as well?

<patrickdengler_> Not animateTransform but any *new* element (sorry)

<ChrisL> scribeNick: ChrisL

Milestones and Planning

<ed> http://www.w3.org/Graphics/SVG/WG/wiki/SVG2_Planning_Page

heycam: need some target dates

<ed> http://www.w3.org/Graphics/SVG/WG/wiki/SVG2_Planning_Page

<heycam> http://dev.w3.org/SVG/profiles/2.0/reqs/publish/

heycam: was supposed to add stuff back into the wiki page from the separate document

ChrisL: and people were to add explanation for the accepted items

vhardy: whatabout use cases?

heycam: should be in the wiki page
... if its useful we can continue working on a /TR requirements doc

ChrisL: want oto avoid work that is of no beefit. wiki page is more easily updated

ed: easier to update wikipage

heycam: easier to read a published doc though

cyril: we have 90 left to do
... out of 180 at the start
... wereat 40% after tpac

vhardy: think its a big effort to migrate to a new document

ChrisL: people do not read reqts top to tail, they drill down on a particular issue
... still feasible to do by end of this f2f?

cyril: we have 2/3 by scrollbar

birtles: doc gets longer as we edit

ChrisL: earlier ones were more vague and we took a lot of time to understand the comments. later ones are specced features from 1.2T so its either in or out
... next, whatdo we have in fpwd - lots or a little (ipp implications)

<heycam> CM: the other question is whether we go forward with this grand refactoring of the document that we talked about before

<heycam> … I think it would be better to not do that upfront

<heycam> … instead we just add the features/changes according to the existing structure

<heycam> … so we can get drafts published with them

<heycam> … rather than having the first couple of working drafts just consist of reorganisation of existing text, with no other changes

ChrisL: would prioritise documenting new features over refactoringexisting ones
... patent exclusions are at fpwd and last call, and are per-feature
... so its better to at least list a feature even if the spec is incomplete or the syntax not stable
... want to see tests early

heycam: so when new feature is added, add tests as well

ChrisL (explains how automatic test assertion annotation is added)

ed: good to get some tests using the new harness, to see how it works

ChrisL: we need to decide link mechanisms. WOFF used this new mechanism (but the files were mostly html)

<scribe> ACTION: chris to continue to work on test suite linking mechanism for new harness [recorded in http://www.w3.org/2012/01/12-svg-minutes.html#action06]

<trackbot> Created ACTION-3212 - Continue to work on test suite linking mechanism for new harness [on Chris Lilley - due 2012-01-20].

ed: we can update the milestones to add this testsuite item

<cyril> http://test.csswg.org/

birtles: good reftests use lime green for pass and red for fail

that is not the right link

http://w3c-test.org/framework/

ChrisL (walks through the new reporting framework)

heycam: need stubs for every new feature for fpwd. say 3 months after the editors draft
... so July
... then another in October for TPAC

vhardy: who can lead the spec editing effort

cyril: needs someone who already has ssh mercurial access

ed: tav has done most of the editing so far

heycam: last entry depends on the feature owners

vhardy: in reqts we have big sections then a lot of tiny ones

heycam: as we go through the list we can allocate owners

vhardy: we need to triage the requirements to prioritise the most important ones

heycam: we did not do that, so we accepted a lot of requirements some of which might not be for 2.0

vhardy: need priority on all the accepted requirements

cyril: once we have commitments for the popular features this will lead to priorities. if no-one commits its low priority

heycam: and if that is loo long we can cut it

cyril: up to individuals how many new features to take on

vhardy: priority needs to be decided before fpwd
... march is better

<scribe> ACTION: erik to remind people early March 2012 about prioritisation [recorded in http://www.w3.org/2012/01/12-svg-minutes.html#action07]

<trackbot> Created ACTION-3213 - Remind people early March 2012 about prioritisation [on Erik Dahlström - due 2012-01-20].

heycam: looks good for the milestones

vhardy: management nightmare to weigh the priorities

heycam: happy to make a form for that

<scribe> ACTION: heycam to make a form for priorities [recorded in http://www.w3.org/2012/01/12-svg-minutes.html#action08]

<trackbot> Created ACTION-3214 - Make a form for priorities [on Cameron McCormack - due 2012-01-20].

ChrisL: goot time for lc would be after tpac so (with holidays) jan would be good, finished by May/June and CR in August depending on need for a second LC

vhardy: for selecting requirements we could update - things thatare broken, graphical features
... for 2.1 2.2 etc we could put the theme that adds value for each iteration, nt too ambitious each time
... first priority could be graphical features, with animation and dom refactoring left for a dot release

heycam: seem to have a range of killer things that do not fit those themes

cyril: this discussion is easier when we have a prioritised list

vhardy: main point is to limit the scope to what is acheivable

heycam: ok so lets aim for LC in January 2013

getBBoxOf

<heycam> http://www.w3.org/Graphics/SVG/WG/wiki/Proposals/getBBoxOf

heycam: this is getting the bbox in a different elements coordinate system
... often want to do this
... proposal is to add getBBoxOf that you call in the element whith the desired coord system and you apss the element you want the bbox of
... need not be a descendent

cyril: bbox is always axis aligned?

heycam: yes

ed: its like getScreenBBox in Tiny

cyril: (points out an error in the proposal, this and element are transposed

vhardy: do we want a rect or a set of points?

<ed> http://www.w3.org/TR/SVGTiny12/svgudom.html#svg__SVGLocatable_getScreenBBox

vhardy: (draws on whiteboard)

ed: agrees with vhardy, you want four points for a non axis aligned bbox

vhardy: want the points, or an option to get the points

cyril: last step is trivial, min maxso return the points

ChrisL: getBBoxCorners ?

birtles: what will people do more often?

vhardy: adding handles is common so the points would be better there

cabanier: illustrator gives the full bbox

heycam: we don't want to multiply the number of methods unecessarily

cabanier: if you miter corners and transform, miter will go further

ChrisL: recall that we also wanted to have an inked bbox that includes stroke and markers
... so for this one which is geometric, stroke mitering does notaffect it

vhardy: bbox is for dirty areas, editing

birtles: hit testing

ChrisL: for painted bbox we would need the analog of this (get painted bbox in another coord system)

cabanier: if transformed, miters would become sharper

vhardy: we do pre-transform stroking
... illustrator gives you both
... non scaling stroke strokes the transformedgeometry

heycam: so maybe instead give a much easier way to get quads and then transform them to another coordsys

vhardy: like the quad idea

getBoundingQuad and then transform and getBBox on that

<scribe> ACTION: heycam to revise getBBoxof proposal based on this discussion [recorded in http://www.w3.org/2012/01/12-svg-minutes.html#action09]

<trackbot> Created ACTION-3215 - Revise getBBoxof proposal based on this discussion [on Cameron McCormack - due 2012-01-20].

Auto sized image

<vhardy> http://www.w3.org/Graphics/SVG/WG/wiki/Proposals/Auto-sized_image

<birtles> http://www.w3.org/Graphics/SVG/WG/wiki/SVG2_Requirements_Input#Auto-sized_image

cyril: is it auto size or auto scale

heycam: (rapidly corrects proposal)
... its the same rules css uses for intrinsic aspect ratio

vhardy: the initiial viewport is there is there is no bbox
... defaults to 100% 100% and the width is known

heycam: so you dont use any 150 300 ratio

cyril: like the idea to not give w & h if you don't want
... do no t want to break backwards compat

<vhardy> vhardy: at the end of the day, we need to align with the CSS 2.1 behavior.

heycam: prefer 1 over 2 but itdoes change existing meaning
... second one doesn't
... in old viewers, missing attrs gives zero so no display - do that all the time

vhardy: if you have w h and viewbox there is enough to work out the aspect ratio

ChrisL: people depend on missed attrs going to zero, so I prefer the second option

vhardy: common case of images with same width and variable height

cyril: so that leads to proposal 2
... how aboutsizing="intrinsic"

ed: i've seen some (rather complex) js driven frameworks that fill out w and h later and it does not want it to render if zero

cyril: used for image preload

birtles: is the preload specced?

(we think not)

vhardy: to display an unknown image you load by script just to calculate the w and h

1 - two

2 - four

ppl voting 1 can live with 2

resolution: image resizing will use option 2 with sizing="intrinsic" as an attr not a property

---break---

<cyril> scirbe: Cyril

<cyril> scribe: Cyril

<scribe> ScribeNick: cyril

test suite

CM: are we only having ref tests or scripted tests?

CL: they are not mutually exclusive

CM: yes

BB: the scripted tests have the advantage of finding the reason for a failure easily
... you don't have to dig

CM: there is the CSS thing (ref test) and the testharness.js

ED: James Graham worked on making sure it works with our existing test harness
... so it is possible to use the existing test suite if we had script snippets
... but I' mot sure that's the best way to convert the tests
... I don't know if the existing test set is the best test
... Opera is working on publishing the existing animation tests for SMIL but scripted so that we don't have to look at the output
... it has some requirements on the API: pauseAnimation, and setCurrentTime

<ChrisL> http://w3c-test.org/resources/

CC: at TPAC, the break out on testing was explaining how to use the harness

<ChrisL> http://w3c-test.org/resources/apisample.htm

<ChrisL> http://w3c-test.org/resources/apisample2.htm

<scribe> ACTION: Erik to check in an example on how to use the testharness.js [recorded in http://www.w3.org/2012/01/12-svg-minutes.html#action10]

<trackbot> Created ACTION-3216 - Check in an example on how to use the testharness.js [on Erik Dahlström - due 2012-01-20].

CL: the current tests, we only have the image patches in the case where Batik does not pass
... now we want one of all
... and they need to be a lot simpler

VH: that could be a real time sink
... I would like to say that we don't touch the old tests

CL: that's a good comment

ED: apart the animation tests, that should be fine for the others

VH: especially with the new Browser API, is there anything to do screen capture

CL: it has been proposed as modal and for the specific purpose of testing

<ChrisL> but not enabled by default, for security

VH: do we agree to keep the old test suite
... as is and work on a new one

CL: yes

ED: yes

CL: we can copy over if needed

CM: we would need to annotate the old tests for the coverage evaluation

CL: the current annotation does not have the granularity
... that we need
... you need an id and a class, the test points to that id
... I did that with the WOFF spec and it's quite simple
... an id on each assertion and a class for viewers, authoring tools ....

Spec review requests

CM: we got two requests to review CSS specs
... we need to assign actions so that that gets done

VH: a third one is shadow dom

CL: CSS Values and Replaced Content

<ChrisL> Title: CSS Image Values and Replaced Content Module Level 3

<ChrisL> Temporary URL: http://dev.w3.org/csswg/css3-images/

<ChrisL> Title: CSS Basic User Interface Module Level 3:

<ChrisL> Editor's draft: http://dev.w3.org/csswg/css3-ui/

RC: the first one has gradients

CM: UI has pointer events and cursor
... UI is meant to be reviewed by 14th of Feb
... Image Values byt the 17th

<scribe> ACTION: Vincent to review the Shadow DOM draft [recorded in http://www.w3.org/2012/01/12-svg-minutes.html#action11]

<trackbot> Created ACTION-3217 - Review the Shadow DOM draft [on Vincent Hardy - due 2012-01-20].

CM: it told dmitry we will find in the SVG spec, where shadow dom is useful
... I didn't give him a timeline

<scribe> ACTION: Chris to review CSS UI [recorded in http://www.w3.org/2012/01/12-svg-minutes.html#action12]

<trackbot> Created ACTION-3218 - Review CSS UI [on Chris Lilley - due 2012-01-20].

<scribe> ACTION: Erik to review the 3 specs [recorded in http://www.w3.org/2012/01/12-svg-minutes.html#action13]

<trackbot> Created ACTION-3219 - Review the 3 specs [on Erik Dahlström - due 2012-01-20].

<scribe> ACTION: Cyril to review CSS Image Values and Replaced Content Module Level 3 [recorded in http://www.w3.org/2012/01/12-svg-minutes.html#action14]

<trackbot> Created ACTION-3220 - Review CSS Image Values and Replaced Content Module Level 3 [on Cyril Concolato - due 2012-01-20].

<vhardy> ScribeNick: vhardy

<heycam> Shadow DOM spec: http://dvcs.w3.org/hg/webcomponents/raw-file/tip/spec/shadow/index.html

CSS Compositing

<cyril> http://www.w3.org/TR/SVGCompositing/

ag: last time we edited the spec, I started putting in parts for CSS so that the document could apply to CSS and SVG, both.
... I do not remember how far we went, and there is more work to be done.

<cyril> http://dev.w3.org/SVG/modules/compositing/master/SVGCompositing.html

rc: did you send a pointer.

(group looking for the draft).

ag: the wording was changed, for example, on enable-background, to be generic and talk about 'host language' and use SVG as 'an example host language'. The document in dev is doing that migration.
... same for knockout and compop

rc: do you remember why did you add Porter Duff.

ag: filters were not enough because when you use the compositing operators in filters, they include the background twice and so you get an incorrect result. This is why we added the PD operators in the spec.

cyril: I have dug up in previous documents/proposals, one of the reasons was that filters require a filter region, but for PD, you can do it while rasterizing and you do not need a region.
... I'll send a link to a document created by Craig Brown.
... if you have a set of elements that are composited with the painter's algorithm, and you want to make one compositing change on one element, you would have to completely modify the DOM (if using filters).

rc: I would like to see an example to understand.

<cyril> example:

<cyril> <path fill="#00FFFF" stroke="#000000" d="M103,120.5c0-33.137,26.863-60,

<cyril> 60-60c33.137,0,60,26.863,60,60s-26.863,60-60,60

<cyril> you want to modify to the third object to be composed with multiply

<cyril> <path fill="#00FFFF" stroke="#000000" d="M103,120.5c0-33.137,

<cyril> 26.863-60,60-60c33.137,0,60,26.863,60,60s-26.863,60-60,60

<cyril> <filter id="X" filterUnits="objectBoundingBox" x="-50" y="-50"

<cyril> width="100" height="100">

<cyril> this introduces the filter element

<cyril> whereas:

<cyril> <path fill="#00FFFF" stroke="#000000" d="M103,120.5c0-33.137,26.863-60,60-60c33.137,0,60,26.863,60,60s-

<cyril> 26.863,60-60,60

<cyril> is much closer to the original

rc: ok, so the argument is that it is easier to use.

ag: thre is more than an ease of use issue, there are functional limitation, the background is included twice.

<cyril> Note that

<cyril> using feBackground has the effect of the background being used twice – once

<cyril> as an input in the filter – and a second time as the result being drawn onto….

<cyril> If the background is not initially opaque this causes strange effects.

rc: I have proposed to separate the spec. in two. For PD, you never need the background.
... explains the way PD happens within a group and does not effect anything outside the group.

<svg>

<g>

</g>

No matter what the compositing operator is on r3, it only composites with the content of <g> at the time it is rendered. In this example, only with r2. There is no compositing done with r1 when rendering r3.

The rendering of <g> will then composite with r1.

<Anthony> http://www.svgopen.org/2005/papers/abstractsvgopen/

Discussion on enable-background and compositing, looking at the examples in section 8 of http://www.svgopen.org/2005/papers/abstractsvgopen/.

There is agreement that compositing is done between a src and dst and that following the painter's algorithm, the default dst would be the content of the parent group at the time an element is rendered (i.e., the rendering of the element's previous siblings in the parent group). There might be use cases for compositing with the a destination that contains the rendering of the parent group's siblings. In the preceding example, it means that r3 would

composite with either the content of its parent group (r2 only) or, in a different mode, with the content of r1 and what is already in it's parent group (r2).

ag: the reason for the naming of new/accumulate is that it describes the initialization of the group buffer used with compositing.

<scribe> ACTION: cyril to work with Rik and discuss with Alex to resolve whether, for the purpose of compositing only, we need a control to select the destination used for the compositing operation. [recorded in http://www.w3.org/2012/01/12-svg-minutes.html#action15]

<trackbot> Created ACTION-3221 - Work with Rik and discuss with Alex to resolve whether, for the purpose of compositing only, we need a control to select the destination used for the compositing operation. [on Cyril Concolato - due 2012-01-20].

Moving on to blending.

(discussion about enable-background in blending context).

<svg>

<g>

</g>

r2 is multiply

g is is enable-background: accumulate

Agreement that in both compositing and blending, the behavior and controls for enable-background are the same. In both cases, the need is to control how the 'group' buffer is initialized before it is used as an input in the blending or compositing operation.

(demo of compositing, blending and knockout groups in a popular illustration authoring tool)

<cabanier> example of knockout and enable-background: http://lists.w3.org/Archives/Public/www-archive/2012Jan/att-0007/KO_isolate.pdf

RESOLUTION: the group agrees to separate compositing from blending in the specification. The group agrees to add the missing blend modes. The group agrees to have the default blend mode to be 'normal'.

<scribe> ACTION: rik to resolve issues on clip-to-self, enable-background on compositing and knock-out. [recorded in http://www.w3.org/2012/01/12-svg-minutes.html#action16]

<trackbot> Created ACTION-3222 - Resolve issues on clip-to-self, enable-background on compositing and knock-out. [on Rik Cabanier - due 2012-01-20].

- DRAFT -

SVG WG Sydney 2012 F2F day 3

12 Jan 2012

Attendees

Contents