SV_MEETING_TITLE -- 12 Nov 2013

<ivan> hi tzviya, we will dial in shortly

<ivan> but people are just gathering...

<tzviya> ok, thanks

<gcapiel> scribe: gcapiel

<mgylling_> http://www.w3.org/dpub/IG/wiki/TPAC-F2F#Tuesday_12_Nov

mygilling: let's look at agenda

… robin berjon is coming in to talk about html5

… we also have toby coming in to talk about webdrive and the future of testing in ebooks

… other than that agenda is stable

ivan: let's take 5 minutes

… there's been a misunderstanding. community that predates our work by 1.5 years. there's also a mailing. people didn't know what to sign up for. which has created problems

… let's merge the two mailing lists by either discontinuing or redirecting one

… we will merge subscribers and interest group people will be on both

… and listened to by the tracker

… plan to do it next week

… email will go out on this

<mgylling_> unchanged: public-digipub-ig@w3.org : only IG members can send mail, but anyone can access the archives

<mgylling_> ... and public-digipub-ig-comment@w3.org : completely public both for writing, reading and subscribing, will be discontinued in favor of public-digipub@w3.org

ivan: -ig will be used for mailing agenda

mgylling: back to the agenda

Behavioral adaption

… @tzviya should we start with wiki?

tzviya: sent out email about agenda

… a few ideas on wiki that are categorized. let's talk about background ideas and then look at use cases

… a lot will tie back to prior discussions, but there is room for more discussion

<tmichel> Comments from the public to the Interest Group are welcome on the public comment list public-digipub-ig-comment (public archive). should be discontinued and be redirect.

<tmichel> <tmichel> in favor of ...

<tmichel> <public-digipub

… the first issue is establishing vocab for finding markup

… looking at use case of book semantics

… publications use a lot of components, perhaps we want to standardize component names

… examples are different components of chapter title or sidebars with titles

<tmichel> the IG list is .... All Interest Group participants are automatically subscribed to the group's main discussion list public-digipub-ig (public archive).

… case study is another example from book publishing. other parts of publishing have their own uses cases

… when we talk about dublin core metadata we have spec for when we use title versus subtitle. in this case we're talking about chapter title

… benefit will be improved display, usability and discoverability

… use case in front is w3.org/dpub/IG/wiki/StructSem_UC

mgylling: let's see if we're talking about the same thing. if we look at epub the formal name of titles that a book may have is in package.

… in the future the package file might be redundant and we can do it in just the html

tzviya: this is getting more granular

mgylling: we all know that rdfa exists and there are semantics at schema.org that enable some of this

… we need to find weak points where we need to focus

… i.e. what's missing in the replacement for package metadata

tzviya: i could argue with you whether it could be done in schema.org do this level of detail

mygilling: vocabs are shallow right now, but it would be worked with them

tzviya: this use is trying to show the detail

… more details need to be explored if this is to be done widely

… each publisher is doing it differently

mgylling: that is the problem. lack of predictability

… and lack of detail as you say

ivan: one thing that i would warn against is to use the class attribute for this kind of thing like you did here in the the example in the wiki

… i would consider that an anti-pattern

… when i have discussions with guha and why they don't want to use microformats. they said usage of class attribute for vocabs created lots of problems with clashes with other uses

… we should stay away from microformats

tzviya: i understand, but that is how is it commonly done

mgylling: on that topic if the problem we want to get through is the lack of detail and predictability we could go with rdfa

ivan: that is my other point. with my semantic web hat, from a purist point of view if we talk about rdfa or microdata what we do is assign some properties to a subject, which is a URI

… what is the URI for that book instance? or that book publication. we don't have an answer

tzviya: exactly. that is why we have the issue in the wiki

ivan: let's pick up the issue of identifiers at some point in the future. there may be other areas with the same problem. or we use some dpub specific attribute like are used for aria or its

… which is the kind of discussion we may have with robin. decision has to be made or some unknown third one

mgylling: but in terms of this particular use case. maybe we are deviating, but i think we may want to lift from behavioral adaptive, since this is purely a metadata issue

… we need a way for predictablity at the microdata and html level

jeank: going back to the edupub conf. there was heavy use of class attribute demonstrated there. this is a practice that has gone a long time. we need to straighten that out before stuff goes too much further with 100 day initiative coming out of edupub

… a lot of what tzviya is talking about is not just. need to be bumped up to top of priority list

mgylling: when robin comes we will talk about this and you're right that edupub profile is mixing class attribute in. and it's an anti-pattern

… if you look at oreilly they also started with class and they discovered ...

jeanK: oreilly came out it differently than pearson and others. it would be better if we had a schema.org solution. but we don't have that how.

mgylling: oreilly know uses data-type instead because it gets messy

jeank: i heard that tshould not be using data attribute either, because it was a misuse

ivan: that's correct and we had long discussion with html5 people and we were pushed back

gcapiel: bib-ex is looking at this maybe

jeanK: level of detail is not there yet

tzviya: i agree that is high priority

… class is playing a strong role

… i have a few use cases where we need to define markup. i have a few examples around index and notes and media specific markup.

… let's look at the index ones

… http://www.w3.org/dpub/IG/wiki/Behavioral_UC#Behavioral_1

… this is bit complicated because of filtering. once you have categories in place then concept of collapsing cats is not as complicated

… index 3 : http://www.w3.org/dpub/IG/wiki/Behavioral_UC3#Behavioral_3

… this is a bit more complicated, but should be achievable in markup

ivan: i would like to understand from html / css friends pov whether something is missing

… i was wondering if for these problems what is missing in the html/css and other friends? or is something that can be implemented with judicious javascript and it's not a spec issue

tzviya: it can be done today with scripting, but it's complicated and labor intensive. if we had a uniform way to do it and this term indicates a category of indexing like in epub

… then it would be far simpler for creating indexes or collapsing indexes

ivan: from this tech pov this is the same problem as the previous one

… if we have a universally accepted way of adding metadata to the various terms, then you get the addt info to make this simpler. is there anything more than that?

mgylling: i think not?

… this is exemplify why we need additional metadata.

ivan: yes

tzviya: yes and next one is also a markup issue and is doable with css. i have a product that will appear across print and digital and many of us do our markup this way.

… if there we an easier way to have digi products skip print-only section. concept of pseudo selector crossed my mind

mgylling: we could use media, but it would be helpful to use markup. question of why not do this in css?

tzviya: could be done in css. it's just avoiding wrapping it in a span, because it's easier to style than something more concrete than a span

… not sure how do this with css than having to do a span for different formats. it would make like easier.

mgylling: this is a bit different than others. it is an authoring time switching mechanism

tzviya: exactly

ivan: to be honest for that example i think using class is perfectly fine and is the usual usage pattern. so to mark part of the file as print or digital only with css is fine.

… maybe the on the idpf you want to have standardized class attribute names unlike html5. ow, using css class selectors seems fine

mgylling: so in terms of gathering examples of why addt semantics are needed which is where we started.

… book semantics 1 should be about predictability and depth. but others are examples of extending html

… are there others? for example, more direct behavioral imperatives like ibooks:background-audio-track or something like that that are not epub or data type

… but dedicated attribute that specifies nature of something

tzviya: can look into ibooks specific attributes

jeank: we just need to document their custom epub like types. some drive navigational type features. we can pull a list together

mgylling: doesn't need to be complete, but has a fair number of examples of what pub industry is doing and it's a hetero set

tzviya: need to be careful about IP

mgylling: we need to generalize use cases
... need to look at all special hooks out there

… are you two sufficient sources of all this info?

tzviya: we can use more info
... we are working with bisg looking at this.

mgylling: that group will have all this
... to summarize we just need mechanism of examples to make it clear, not an exhaustive list.

… we just need enough to know what's absolutely needed

jeank: those of those outside of epub are doing some unusual heuristics.

mgylling: heuristics instead of publisher provided markup

jeank: it also may be interested to speak with readium, since they are familiar with w3c specs.

mgylling: readium does not contain heuristics, i believe.

ivan: getting back to id issue. is it correct that in all these examples that the metadata is not on manifestation level (e.g. the book on my ipad)
... let's supposed there's an ideal world where isbn are unique. then id to be usable for microdata for rdfa using id. however if it's added to book that i own, then this id doesn't work anymore.

… do we have the easy or hard problem?

tzviya: if i understand correctly, if isbn is unique...

ivan: the isbn would be the subject for all the metadata and that would be kosher because it gives unique id. but that would not work, if i want to add to the instance on my device.

… are your examples on first or second?

… second case would be for use annotation

… we may choose for this specific use case category, we may choose to do it where ID can be done easily.

tzviya: to make it more complicated. if i have a book with 10 chapters, i have one isbn for all the 10, but then i want to sell chapter 1 separately

ivan: not a problem using fragment IDs. we have to define fragments, but not a major problem if we have an id of work

brady_duga: only if id is unique for chapter 7

ivan: isbn for title. problem comes if i want to annotate title on my machine.

brady_duga: you need some unique id, but you're assuming that you're assuming the subset is using the same identifier

ivan: not sure what practice is. whether isbn is good enough

brady_duga: some do not use isbn

ivan: then we have a problem

tzviya: do we need a unique id to sell chapter 7?

… pub will need to be rewired. fragment id creates unique id.

<mgylling_> darobin: you joining us soonish?

brady_duga: questions of how you know ownership

… based on a unique ID

… out of my area

tzviya: can't imagine doing this without isbn

ivan: not sure what practice is with publishers

brady_duga: lack of unique ID is a bug.

… spec should assume that there are unique ID

ivan: one way of moving forward of making problem a bit easier, then we don't need HTML5 extension. not sure whether schema.org has sufficient vocab or epub needs to define it with rdfa without problem is another possibilty

tzviya: libraries do not use isbn. most libraries use loc cataloguing number.

jeank: each country has their own way.

mgylling: ivan in your vision are you saying everyone has to use same scheme?

ivan: no, putting on semantic hat, what rdf says subject has to have unique URI

… we prefer http URI, but urn with ISBN

… it's ok. but if we use rdfa we need identifier. microdata doesn't address this issue.

… because you use itemscope and props

… it's left vague and undef. personally it does not seem like a clean approach and html5 folks will disagree.

mgylling: ivan what is the takeaway here?

ivan: 1st we have to understand the options and robin can help. what does it mean if epub wants to have it's own approach. e.g. epub- or something like that

… tzviya's examples could be done that way

… that's a way of doing it

… if it's messy or if we prefer to reuse vocabs then we can use microdata or rdfa approach and need to check whether vocab exists that covers the needs, which it probably doesn't.

… and it needs to be define and we need to figure out ID

… if we go html5 route, the vocab issue is still open

mgylling: discussing two different things. struct semantics stuff where there is a pretty unison opinion that rdfa is overkill.

ivan: rdfa lite is not an overkill. similar to microdata in terms of complexity

… we need vocab and someone needs to define it

mgylling: idpf has it

… not following. between prior discussions

ivan: it's a mix. indexes for example via the index you def something deep into the contents that could be used out of that role as well

… not clean

tzviya: can i send you an example that you can markup?

ivan: yes, not this week

… but send it

<JeanK> +1 on Tzviya's request for Ivan's RDF vision!

mgylling: taking a 2-3 minute break
... we could use epub- for class attributes

<ivan> Guest: Robin (darobin) Berjon, W3C

mgylling: welcome robin. timely we were just talking about its 2.0

… robin is one of the editors of the html5 specv

… basically the problem cluster we have been wandering around is about metadata and behavior on top of html

… situtation is not just about epub being stuck in html serializations, but also how to enhance semantics. epub uses epub:

… oreilly is working on authoring time schema using class attribute, but abandoned in favor of data type attribute

darobin: that's wrong

mgylling: pearson is doing a mix.

… we would like to quickly identify a solution that is unified and serialization neutral

… collection that works in xhtml and html

… and looks and behaves the same way

… it would be interesting to look at its 2.0

darobin: so i'm not super familiar with ITS details.

… what they have done is taken a bunch of attributes with its- as a prefix. the validator considers it as applicable specification and accepts it as valid html

… this mean that the same approach that the same approach could be used by epub

… in the same sort of spec for one or more attributes that would function the same

ivan: what does an acceptable / applicable extension mean?

darobin: we don't want to list everything that a user can support

… people should be free to mix stuff up

… based on what the user agents support

… if epub- or book- or digipub- is implemented then it's fine

… by html working group standard it's perfectly

ivan: does the html5 validator need a stable document spec published by somebody

darobin: technically we don't need anything. we don't give out stamps of approval

… that being said, if you want it accepted by w3c validator, one way to convince them you need to have a royalty free public spec

ivan: if we produce a spec with a set of attribute epub-xyz and later epub-aaa, then it's not a problem. we just need to specify epub-

… if idpf or this group publishes a note of the way we want to go. is that good enough?

darobin: can't vouch for validator team, but that seems to be what is accepted

ivan: does it need to be much?

darobin - if you remove the boilerplate, 2-3 pages

… doesn't have to be complicated

tzviya: what about extensibility of this list?

mgylling: that's the values

ivan: what i'm undestanding is that the validator accepts all attribute epub-

darobin - no, they don't like wildcarding

… they check syntax. and it depends on the spec

… and the value space that you are defining

… attributes names need to be fixed

… because in it's hard to check partial names

ivan: that's doable

darobin: you can always add attributes

… you just need to coordinate as long as it's not an attribute a month

mgylling: epub-type would be primary one
... from idpf perspective, we want to move fast.

… what does w3c suggest?

… it's a note. which only takes about a week once doc is done

darobin: could be 24 hours once publishing happens

ivan: not a big deal

mgylling: should this happen in w3c or idpf space?

ivan: idpf is probably a better known place because it's not seen as much as an outsider

… doing it only as w3c doc is not a good idea

<JeanK> ++1 to the idea of a jointly published note from both the IDPF and W3C...

ivan: we will find a way

mgylling: let's talk about pagination dom events

… let's summarize quickly

… both tzviya and jeank are continuing to gather examples of custom attributes for behavioral extension or suppression

… we did ask to move title stuff into metadata

… we have identified a path for epub- space that we will pursue

mgylling: what about identfiers conundrum?

ivan: it seems for this category of use cases the approach we discussing with darobin is probably the way to go, thus ID discussion is not necessary for this issue, but will come back for other topics

… let's push it to annotation people where it's a major issue

<tzviya> we lost the phone line

<JeanK> Tzviya and I are!!!

<duga> We are trying to get back on

ivan: [replaying discusion]

dom events, pagination

mgylling: should we quickly cover dom events and pagination

brady_duga: speaking of pagination. dave cramer is working on issues around it. questions around pagination really means. when we take a html doc and break into pages. what is a page and what pages should be visible. ordering. how do we know a page has loaded. we talk about styling, but not from a dom perspective.

darobin: good questions. lots of diff answers. if entire book is a doc or something large than page. page is viewport.

… does not give you page number

brady_duga: not saying everything is defined. some pieces are there in css, viewports in css object model. recommend creating new events like next page, previous page. this is easy.
... viewport is not quite good enough.

darobin: don't use xquery for this

brady_duga: don't have answers to this stuff and different browsers doing diff things

darobin: don't do this beyond for making content interoperable.

… for ex mobile alliance, didn't have a good definition of what a mobile device was

… laptop? time was lost in def

… agree a subset of questions need to be answered, such as page change and styling for pages.

… you may not need all of it for good content

brady_duga: maybe the answer is there is no such thing as a page. right now it kind of works, but is painful. would be better to do at browser level for performance

ivan: need to understand way forward by seeing what various devices do

brady_duga: webkit kind of has this.

… but not exposed

… not exposed to JS, so you can't go to a page

… internal trick

ivan: we need the same thing we need for the metadata stuff. let's collect info what's done today and what's painful.

… so we can understand the level of abstraction needed.

mgylling: we need use cases from user point of view and what pubs need.

to trigger actions

… and various things actors need to do

ivan: readium should give us info where problems where

<tzviya> q

<tzviya> +

tzviya: term pagination means many things and it's confusing

… we use it to mean parity with print book and also talking about it from a visual standpoint

… for example dealing with widows, etc.

… brady_duga is talking about event triggers

<JeanK> +1 to Tzviya. We need a taxonomy.

… let's not forget that the same issue doesn't happen in just browsers

ivan: so operationally, let's say this group produces a doc and we give you doc.

darobin: depends on what's on the doc. it should not be about styling.

ivan: we have a separate doc on styling issues

darobin: if it were reqs for html, js and dom, then it could happen in html working group

… not sure it can be pulled of by darobin on his own

… html5 official req still needs to be shipped

… area i'm in interested in

… and others too

ivan: we knew this is the way this should happen

darobin: we do see this needed in browsers too

brady_duga: we have reading systems in browsers too

ivan: do we do this with dom4?

darobin: yes. dom4 is about fundamental trees and is extensible

ivan: let's take 15

<fjh> /me will probably drop of in 30 minutes or so

<tmichel> scribe: tmichel

security

digital publishing security issues

digitral books benefits (low costs, portability, richer content, easy to distribute, ease of use, to modify

but several challenges and issues ...

distribution content ...

publisher create content and provide content to 3rd party stores ...

Users buy content from 3rd party Stores or Market place

scribe: Users can purchase content from publishers site ...

There are security at each of these level.

ant each level from moved publisher to store / 3rd party stores content purchase

<scribe> New slides: publishers related security issues

<fjh> is there a need to distinguish attack of individual taking content for personal use versus taking content to publish in volume, cannot the first be handled by pricing?

UNKNOWN_SPEAKER: during content development

right management (multiple right management, each with strenghts and weakness)

transmission/distribution

making sure the digital content is securely transmitted.

Next slide Client side issues

encryption on Device dpub books are always encrypted on users devices

securing the Licence/ Encryption Key on device

Access control (enforcing publisher policy like cut and paste, disable printing) ....

or limitrd number of devices content can be opened on.

Interoperability issues (how to play content from different publishers using security technos

frederic: republishing issues ?

Sharad should present this at the end of his presentation

<fjh> my question is whether we can make sure we frame the problem correctly, since otherwise there is much detail

NExt slide: Un authorised Copy and Distribution

<fjh> is it possible to focus on large redistribution rather than individual?

<fjh> otherwise concern about usability, convenience, accessibility etc?

<fjh> really just a question

Offline Consumption (paid ebook for reading off line)

and copy and un-authorised distribution

prevent un-authorised copies

Next slide: browser support for digital books

<fjh> what I think I'm hearing is that DRM requirement depends on target market segment, and that text books may be different than others

digital content can be consumed from publisher's website via browsers without downloading onto a local device

which triggers issues.

next slide Next steps

provide input, develop detailed Use cases

privicy issues.

Mgylling: what are the new things you want to provide that are not existing in DRM ?

<fjh> I suspect users don't want DRM at all

providerss and publisher have each their own DRM system

Ivan: industry does not have an agreement. So there is a huge variety of systems for DRM

<fjh> watermarking offers a mitigation of bulk copying but does not disturb legitimate user

Sharad: So many solutions that are not all working

therfore should list use Cases ...

Ivan: there are major technologies to encrypt and signed ...
... on the DRM side I have difficulties to deal with this ...

Sharad: it is not only DRM but also security issues
... not trying to give solutions ... but trying to collect UC and requirements.

Then, see to who we should input these requirements.

And maybe technology already exist.

ivan: the UC metadata we discussed this morning for example are easy to have consensus within the group. It may be more difficult to get such consensus about the validy of these UCs for security and DRM

mgylling: this iw exactly what happened in idpf when we started collecting these kind of UCs.

We couldn't achive anything in this field.

Sharad: happy to do some more work if needed.

ivan: move issues to look at like security and privacy issues. Do they apply to epub.

<fjh> DRM is a different issue than security issues, agreed

<fjh> We definitely should understand such broader security issues, as mentioned in my email

privacy issues about ereaders collecting my behavior abouthow I am reading my books.

<fjh> +1

scribe: might be really relevant to look into security and privacy

sharad: will work with frederic to work on UC and requirements on security and privacy issues.

Mgylling: Any one to repport on EME ?

ivan: current psycho drama ;-)
... for book readers I guess EME would not be the right approch.
... need to have a secure envirement within the browser.

mgylling: anymore issues ?
... maybe we should break for lunch ...
... After lunch we will discuss security issues

<duga> test

<duga> scribe: duga

Accessibility

Guest: Jason White

<ivan> Guest: Jason White

Jason: intro - lots of experience on standard, a11y
... joining us to discuss a11y in digipub
... will help us pilot through other WGs

mgylling: This IG is looking at everything, a11y is high priority
... in ebook space still trying to get to a point where we can maximize web a11y tech
... have a history where the entire web access stack is missing
... want to enable it everywhere
... Are there differences in priority for the web vs books
... for instance, web browsing vs a university student studying
... need to point out differences
... eg need to have efficient, accurate speech synthesis
... 2 themes - 1. how do we content and impl on aria bandwagon
... 2 - are there any gaps?

gcapiel: will run through unordered laundry list
... then lets prioritize
... How do we broaden the range of perspectives in the group?
... need to cover all areas of assistive tech
... Would like input on who we should pull in
... Problem areas - SVG
... lots of features for assistive tech, not widely adopted
... May not support everything we need
... descr attribute can't have rich text

ivan: What about html via foreign object?

gcapiel: not supported by screen readers
... need to drill down on use cases
... image descriptions, epub 2 limited to alt, epub 3 has no more support, 3.0.1 has Aria described-at
... no actual guidance/ best practices, or semantics of the document
... might want short description, long desc, etc
... These are new areas we need to define some semantics

ivan: There are a11y problems with SVG - is this epub issue, or general svg

gcapiel: everywhere, but more pronounced in edu
... as it moves online, may be on the same footing. Older use cases may not account for modern uses in edu
... for edu, in epub, since it is packaged it is harder to update once released
... Need remediation tools to fix epub that is missing a11y features
... may want to use annotations to implement that
... MathML fallbacks is particularly a problem in epub
... hybrid epub 2/3 files use nav features but don't make content epub 3 (so, no MathML)
... need to explore mathml fallbacks (alt text, images)
... none really good enough
... even with mathml, works well for visual representation, but not great for aural
... some people want nemeth (??) braille conversion from mathml

mgylling: nemeth braille may be US centric

jason: mathematics codes not standard
... mathml to braille conversion hard, need to convert to local braille math representation

gcapiel: will a publisher really account for every math code? Need to make sure easilly adaptable

guest: Andrew Kirkpatrick, Adobe

gcapiel: CSS braille not gone very far. True?

Jason: Started end of the 90s, didn't go very far
... issues around grid for east Asian typ0graphy, lots of questions on if possible to make a braille for that

ivan: East asian typography just picked up in general, so maybe it can be looked at again

mgylling: Following the CSS group, has not seen this in 12 years (css braille)

gcapiel: looking at runtime braille conversion

andrew: useful when we have haptic displays?

jason: maybe now is the time

mgylling: good time to bring it up, as epub is looking at alternate renditions. Might make sense to look at braille as alternate rendition
... braille is so hard, might be hard to do in automated fashion, so hand crafted is good

andrew: and there are legal reasons in the US

jason: And stuff like tables is really hard to automate
... back to svg
... aria support getting in to svg
... area states and roles can be placed in an svg document
... no aria changes for svg, but could be worked on if there are specific use cases
... could associate aria role with part of svg document, then make accessible ui
... don't know if anyone is working on it

mgylling: new member of the group looking at it for STEM (may include inforgraphics)
... this group could propose extensions to aria
... mark Hakkinen is the new member

jason: knows mark and knows he is interested in aria extensions

gcapiel: css speech - don't know what we need
... implementation lacking

jason: where are we on the spec?
... was in 2, going to drop in 2.0.1

mgylling: was revived in epub by daniele weck
... it is at least a WD in CSS, but has gone nowhere
... is marked as at-risk by css wg

ivan: why?

mgylling: no implementations?

jason: protocols and formats wg interested in people reviewing and getting engaged in what is happening in css
... could be a convergence of interest
... pay attention to and contribute to css

andrew: they are trying to follow css, but it is huge and hard to follow

gcapiel: TV says this group should file a bug against chrome, since it is lacking css speech

mgylling: there is also the issue of pronunciation dictionaries
... there are specs on it, but they are not part of OWP
... epub 3 references them, but there are no implementations

ivan: haven't heard any epub specific issues
... all OWP issues that we happen to share
... this group should concentrate on whether there are really epub specific problems
... we can't comment on css in general

mgylling: as gcapiel points out, some of these are more critical to dpub
... the evolution of the epub ecosystem depends on these things

duga: similar to hanging punctuation - more important for us than OWP in general

ivan: are there features in the specs that are really missing?
... lacking implementations are not something we can deal with

gcapiel: well, there is braille

ivan: yes, css braille is one we should look at

jason: a range of a11y issues haven't gotten enough attention
... need to get people involved in standardization of these issues

ivan: if I take an ebook reader, there is a lot there not in a webpage (toc, etc)
... are any of these a11y features like those ebook specific features?
... are any of these things ones only we care about?

gcapiel: biggest one is annotations piece
... braille, annotations, ...
... So spec for synchronized reading, but there are implementations

andrew: word or line height seems like UA implementation detail

ivan: elevator pitch is look at publishing industry needs, look at w3c tech, see if there are any holes
... yeah, it a ua guideline issue, but that means the guidelines are not sufficient
... these are the things we should try to collect

gcapiel: in that vein, web speech is another one
... not going to cover the other ones to keep it focused
... use cases focus - braille, annotation, maybe web speech

mgylling: infographics and assessments are lacking

ivan: In an ideal world we should look at everything and see what is missing
... we have 2 or 3 items - are these all of them?

gcapiel: probably not - need new members
... wonder what high priority items are for pubs

andrew: uag approaching last call
... atag in the same situation
... stuff like dictionary popups, could be in content, could be in UA

ivan: seems like it is all UA at the moment

jason: is there such a thing as a full html implementation of epub in JS?

mgylling: yes

jason: that means all the OWP tech is in place
... though may not be accessible

ivan: shell around web UA, need to look at that shell not just browser
... do ereaders use canvas?

mgylling: there is a japanese reader that is entirely implemented in canvas
... because of crappy typography

ivan: canvas is an a11y nightmare

jason: yeah, canvas is nasty for a11y

andrew: there is work around shadow dom to make canvas accessible

mgylling: in terms of the list of concrete perceived gaps, what is the list?

gcapiel: assessments for aria, infographics, braille (css), web speech (UA)

jason: that includes css as well, right?

gcapiel: Huh?

jason: you referred to css speech

mgylling: need pronunciation lexicons

ivan: mathml?

gcapiel: mathml semantics, yes
... bridge gap between content and presentational mathml

mgylling: good luck with that
... is there work on this?

gcapiel: I don't know

mgylling: what is the state of this?

gcapiel: it's pretty easy to write crappy mathml

ivan: mathml is one of the issue that is more important to books than owp

gcapiel: and there is the fallbacks issue

mgylling: do we care how we generate good mathml?

andrew: need a tool to generate the good mathml and send it to TTS
... design science has some success

gcapiel: but it has limited scope (IE only)

jason: suppose there was a techniques document for constructing content mathml. Is that a good start?
... or do we need to change mathml to get halfway?

andrew: might accept it as an advisory technique, but there us still a big gap

jason: disagrees. wcag conformance requires relying on certain technologies

jason: huge challenge to get eberything working across various APIs and OSes.
... either need to enable web tech to act as a11y tools, or we need to get everyone to implement all these specs

mgylling: in terms of the short list, how does it look?

gcapiel: will need help on assessments

mgylling: mark hakkinen to help with that

gcapiel: can look at infographics and web speech
... has some mathml knowledge, but may need more help
... need to go through guidelines and see what is missing

mgylling: when is UAAG up as a draft?

andrew: atag is soon or already happened.
... before eoy, both of them

mgylling: LC?

andrew: one is CR, maybe

mgylling: doesn't matter, we can adapt

gcapiel: web API for braille support is a new one

andrew: atag was CR on nov 7th

jason: mathml is a significant issue
... makes content inaccessible without it

ivan: it is one of the most frequently mentioned issues

gcapiel: mathml barely comes up in UAAG guidelines

andrew: UAAG was LC nov 7

jason: what can the guidelines say at this point that will help?
... need new tech, not guidelines
... whatever the solutions (OS, etc), we need tech

andrew: uag doesn't mention mathml explicitly

jason: need the minutes

<scribe> ACTION: ivan to get jason the URI for the minutes [recorded in http://www.w3.org/2013/11/12-dpub-minutes.html#action01]

<trackbot> Created ACTION-11 - Get jason the uri for the minutes [on Ivan Herman - due 2013-11-19].

<mgylling> scribe: Sharad

Testing (with Tobie Langel)

<ivan> Guest: Tobie (tobie) Langel, W3C

<tobie> https://www.w3.org/2013/Talks/1112-testtwf/

tobie: w3c is so far testing spec is implementable, less you test, easy to move spec forward
... TV industry is getting involved with HTML5, mobil & publishing industry is also moving towards HTML5
... lots of new spec, interactions with new technology thus lots of issues. new effort started to improve testing
... scope is large, # of test is missing, approx 25 - 30 Man years of engineering needed
... how do I test, how to run them, how to gather results. happy to answerrelated questions

mgylling: like to know architecture, couple of open source systems available. testing in epub is strictly limited

tobie: 30 - 40 specs for mobile, TV inudstry. 30 o them may be required for publishign industry
... it will be useful for everyone to have implement these specs
... missing is solid base that developers can use acrossdevices
... we have we have a # of different kinds of tests, most common is JAVA tests
... 2bd kind if test are reference test, these ar layout tests

<duga> IDL

tobie: tobie: special kind of tests: IDL test. automatically generate tests, also called signature testing
... web driver tests; completely mimic a user, navigate to a page, take screen, itake nput

get rid of manual test, replace them by web driver tests

scribe: web drivers don't do testing outside of web viewport;

ivan: is the implementation part of webkit or part of web core?

tobie: i am not rigth person to ask technical details

ivan: publisher will want produce typography nicety

duga: never have hanging puctuation; comma appearing outside of line

ivan: 2 apge side by side, they would not be equalize, today does not exist in CSS
... typography tests need to be checked by human. running test harness automatically wil lbe major help

tobie: cost of test automation will be sometimes more than benefits

mgylling: talk to Readium foundation, developing opensource reading system
... do we good online intro to webdriver?

tobie: no
... nothing about driver structure

ivan: can look at web code to understand?

no funds to suport this effort, this is going to be another issue

ivan: how is it related to validation project?

tobie: completely different scope
... looking for options to scale

ivan: start the process here and then show readium

<dauwhe_> Bert and I have proposed a breakout session for tomorrow about paged media etc.: http://www.w3.org/wiki/TPAC2013/SessionIdeas#Towards_paginated_layouts_for_.28e-.29books_using_Web_technologies

- DRAFT -

SV_MEETING_TITLE

12 Nov 2013

Attendees

Contents

Behavioral adaption

dom events, pagination

security

Accessibility

Testing (with Tobie Langel)

Summary of Action Items

Scribe.perl diagnostic output