SVG Working Group Teleconference -- 15 Aug 2013

<trackbot> Date: 15 August 2013

<TabAtkins> Dear SVG: Please kill your namespaces faster. Love, Tab.

<krit1> TabAtkins: we would need to add the SVG elements to the HTML parser

<scribe> scribenick: nikos

<TabAtkins> krit1++

<TabAtkins> But really it's the xlink that's killing me today.

<TabAtkins> Which requires only local changes to the language.

Publishing and renaming CSSMatrix

<heycam> Removing xlink is something we can do without a lot of the other stuff.

<cabanier> I thought we already agree to killing that off

<krit> https://dvcs.w3.org/hg/FXTF/raw-file/tip/matrix/index.html

<TabAtkins> Yeah, but it's not edited in yet. ^_^

<krit> http://dev.w3.org/fxtf/matrix/

heycam: we haven't published this yet?

krit: correct
... 2 issues
... first is name. CSS WG doesn't want CSS prefix

heycam: dictionary ones don't matter since they're not exposed in script
... interfaces intended to be used outside probably shouldn't have css in the name

krit: suggestion was either TransformationMatrix or TransformMatrix
... I don't have others

heycam: if we named it Matrix, there might be a risk of conflicting with scripts that define Matrix already

krit: we also have conflict with other working groups
... was discussed on fxtf mailing list
... was suggested to name it Deprecated Matrix

heycam: ugly name!

cabanier: web gl want to have something more primitive
... not useful for authors. The reason we are doing this is to make operations easier for them

heycam: I'm not convinced by argument that we shouldn't have it becaues JS authors can write something that can run faster
... browsers can implement with native js behind the scenes and avoid the overhead

<TabAtkins> I'm down with TransformMatrix.

krit: because Matrix is so generic, I don't think it's a bad idea to have a prefix

heycam: you said CSS WG has already resolved to publish? what was their final view on name? have an issue in the spec?

krit: we need the name for the url though so it's problematic to do that

heycam: I think it would be fine, even if we call it CSSMatrix or TranformMatrix to have the url as just matrix
... I'm fine with publishing and including an issue

krit: CSS WG suggested SVG WG takes care of publication

heycam: couple of other issues open in the spec
... 1. w part of the point

krit: the main is whether we use float or doubles
... Mozilla wants floats. Apple wants double
... difference is 10 times on mobile
... javascript just uses double
... so it's just the internal calculation results
... currently I am using doubles in the spec

heycam: I don't think we need to resolve that before publishing
... what is the plan for replacing SVGMatrix with CSSMatrix?

krit: The whole point of this is to replace SVGMatrix

heycam: so we'll just use CSSMatrix. Won't have any inheritance.

krit: It's backwards compatible, except the new matrix won't throw exceptions as SVGMatrix did

heycam: I also remember roc was concerned about the matrix object being used for 3d and 2d
... meaning you need the extra components

krit: I think he's ok with it now

heycam: I'll check with him

resolution: Publish first public working draft of new matrix specification

<scribe> ACTION: Cameron to organise FPWD publication of new matrix specification [recorded in http://www.w3.org/2013/08/15-svg-minutes.html#action01]

<trackbot> Created ACTION-3517 - Organise fpwd publication of new matrix specification [on Cameron McCormack - due 2013-08-22].

Progress on subregions hinting in SVG Filters

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

krit: Erik put together an implementation matrix
... We need to specify exactly how subregions are calculated
... I would like feedback on the table
... especially where we have question marks
... we're not clear on how to calculate subregions for some filters at the moment
... it's important that the filter region doesn't clip things you don't want clipped
... a lot of people don't understand the 10% margin

heycam: I never liked it. It's not usually what you want
... are you suggesting to remove filter region primitive attribute?

krit: no
... if they are not present then the user agent should calculate automatically

heycam: so changing the default

krit: in a way that fixes content

heycam: I'd be in favour of that

krit: so we need to specify what unbound? is for each in the table

heycam: for filters like feFlood or feTurbulence that conceptually have infinite output, what would we define apart from 'implementation needs to work out how much of the final rendered element to keep'

krit: we could specify 120%. it's not nice though

heycam: can't we just state it needs to be large enough so that it doesn't affect filter output?

krit: If you use feFlood and it will fill the whole SVG file.
... could use the viewPort

heycam: feFlood is a special case though
... why can't we say something like 'minimal area based on how that filter primitive is used in the filter chain'

krit: Need to specify that content looks equal on all implementations

heycam: I think that can be achieved by saying the filter primitive is big enough so that it doesn't look different
... when the element is rendered, there will be some clipping at some point
... then you can determine the smallest rectangle based on how it is used in the filter chain
... we can require UAs to generate the filter primitive at the required size

krit: I agree. feFlood is still the problem though

heycam: If you didn't have any other clipping, then feFlood would fill the viewport

krit: Would have to check about breaking content

heycam: do you feel we need to specify how to calculate the area for that primitive?

krit: I think that's a good idea
... to ensure consistency between implementations

heycam: I'd be fine if we had something like that written down. Do you want to do that?

krit: yes

<scribe> ACTION: Dirk to specify how to calculate automatic filter region for unbounded primitives [recorded in http://www.w3.org/2013/08/15-svg-minutes.html#action02]

<trackbot> Created ACTION-3518 - Specify how to calculate automatic filter region for unbounded primitives [on Dirk Schulze - due 2013-08-22].

Collapsing CSS Filter functions and the problem of color clamping on

intermediate results

krit: a lot of filter effects short hands can be represented with a color matrix
... ideally you can collapse the filter chain to a single operation on the matrix
... but we currently define that values should be clamped at each stage
... do we want to keep current behaviour or can we relax the specification to allow optimisation in this case?
... it's a balance between performance and ease of authoring

heycam: is there a benefit to clamping at each filter? is that something authors would need or want?
... there are similar thinks in CSS with color values
... I can't remember where exactly but it talks about allowing values to go out of range and clamp in the final step
... so there might be precedence

nikos: I think it would be confusing for authors to get different results depending on whether a shorthand is used or not

heycam: my feeling is that we should allow intermediate results to go out of range and clamp at the end

<krit> :D

krit: initially I was for clamping at the end. Now i'm not sure. There are cases where you can guarantee that you don't need to clamp
... so I support clamping at each step now

heycam: I can't see an author use for clamping each stage
... if there was one I think it would be better to have explicit control to opt in to that

krit: there is another case. contrast 200% then another filter primitive with contrast 50% and they wonder why they don't get the same result
... it's a corner case though. I don't think it's actually useful to do that.

heycam: I can imagine cases where an image has colour values that are almost fully bright and you apply a brightness filter and you want to clamp
... I'm not sure which I prefer

krit: I think we should keep the clamping for now

shepazu: I've seen some pretty crazy stuff come out of authoring tools

heycam: I'm not sure how much performance benefit there is from collapsing down to a single operation on a colour matrix
... I think implementations often slow down with filters because they are creating lots of off-screen surfaces
... here that shouldn't matter as you can work on the same surface

resolution: maintain clamping colour matrices for filter effects at each step in the filter chain

Geometry API and Accessible SVG Community Groups

shepazu: there are 2 community groups you guys might be interested in

1. SVG accessibility

<shepazu> http://www.w3.org/community/svga11y/

<shepazu> http://www.w3.org/community/svga11y/2013/08/15/roadmap/

2. geometry API

<shepazu> http://www.w3.org/community/geometryapi/wiki/Use_Cases_and_Requirements

<shepazu> http://www.w3.org/community/geometryapi/

use cases and requirements doc (in progress)

SVG accessibility is unclear. There's few normative requirements, a lot comes from the doc that Charles wrote about mapping and authoring guidelines

shepazu: very old. refers to SVG 1.0
... After talking to accessibility experts. The state of SVG is ok, but there's a lot of tests that need to be run
... so the whole point of this group is to write tests and write requirements
... we're going to have a workshop/hackathon

heycam: sounds good. for me I don't feel I have the knowledge to know what we need to change in SVG

krit: who would attend the workshop?

shepazu: there's about 16 people in the group now.
... we want: 1. Accessibility experts who can help us test. 2. People who use accessibility features. 3. SVG people
... and at least one 'testing' expert
... I gave a presentation about accessible data visualisation and people were really interested
... that's where this came from
... first step is to make a bunch of tests, run them and create an implementation report

nikos: what's the make up of the group. Who do you need?

shepazu: we have lots of interested accessibility experts.
... they should be able to find people who use accessibility features
... there's two or three people who are pretty good with SVG
... would be good to have Rich in the group
... looking for someone from Adobe
... I'm looking at September for the hackathon
... October at the graphical web would be good. Cyril, Tav, and Nikos will be there
... so we can chat about that and try to organise a time that is good for everyone
... The other group is the geometry API community group
... the goal there is to come up with a geometry api (SURPRISE!)
... we want to meet use cases for SVG primarily. But if we satisfies use cases for other web tech that's ok.
... I want to be able to do things like calculate intersection of shapes

cabanier: Dirk and I signed up

shepazu: If we can come out with a simple set of use cases and requirements that we might include in SVG 2

krit: should get some SKIA guys in the group

cabanier: is this part of SVG or is it just something that interfaces with SVG?

shepazu: I guess you could interact with other web tech

krit: it would be great to have a document that everyone could reference

shepazu: it would be good to have more people in the group to define the use cases that are important
... it would be good to have a matrix library (SKIA has), point and vector maths
... need to drive requirements on those
... in conclusion. It would be good to get people into the groups to drive the conversation and keep them going

- DRAFT -

SVG Working Group Teleconference

15 Aug 2013

Attendees

Contents

Publishing and renaming CSSMatrix

Progress on subregions hinting in SVG Filters

Collapsing CSS Filter functions and the problem of color clamping on

Geometry API and Accessible SVG Community Groups

Summary of Action Items

Scribe.perl diagnostic output