W3C Annotation Workshop, 2014-04-02, San Francisco

03 Apr 2014


Ivan Herman, Doug Schepers


<ivan> scribenick: scribe

Doug: There are annotations which are inline and out of band
... What should exists?
... We have some goals... decentralizing
... - Interoperate, distribute
... I think the web deserves to read-write
... How do we go from out of band to inline? - anchoring
... We need great compents
... I've got some ideas
... ... DOM events target the data
... Existing event targets are DOM elements.
... Selection psuedo elements

Can we describe a selection and style it?

scribe: Without modifying the DOM?
... Discovery elephants
... Where do we get annotations - feed discovery
... Where do we publish them?

<nickstenn> s/elephants/endpoints/ (!)

scribe: -locally - web
... The need to be practical and solve problems for developers

Randall: Fred "How would we get a produce like this if it is not on the market?"

<nickstenn> the question was "How do we attach an event.dataTarget to DOM events if the data is not in the markup?"

<nickstenn> and hopefully Randall can explain his answer when he turns up -- he mentioned "components" but I didn't really catch much more than that

Existing Annotation Systems

Speaker: Chris

Chris: Code errors and Break points
... The text that you're wring in annotating
... UX Of anntation -REad text -Notice the annotation - Continue reading or decide to read the annotation
... Example screen reader features - Announce when entering/exiting annotation
... We've been doing braining and are trying to improve features for the screen reader

<azaroth> Definitely looking forwards to discussing accessbility with gcapiel and Chris (and others!) over the next few days. Need to ensure that there's a good translation model between aria and distribution mechanisms

<nickstenn> Randall: do you fancy trying to give a one-line summary of your answer to the question about event.dataTarget (and where the data come from)

<nickstenn> *comes

Chris: We have starting points that will hopefully improve screen reading

<KevinMarks> Randall: I'm assuming you specify the dataTarget by reference to the text with markup stripped - the quotation being the annotation target

How do you preserve the annotation?

<Randall> data would either come from application code, probably dispatching (or re-dispatching) an event and filling this field

scribe: We should work to make sure there is a methods that exists in pdf for users

<Randall> I had been thinking of using oa:target inspired data

Speaker: Anna

<Randall> so, perhaps for a text selection, this would be simply the text markup stripped

Anna: Annotations are use throughout education
... Annotions are in musuems, such as 3d reading

<Randall> but we might do better. for instance, maybe it's a target object with its source set to the context derived from microdata and a text selector with a quote, prefix, and suffix

Anna: Annotations on animal learning
... Et

<tantek> Randall - do you have a URL to your slides?

<Randall> https://docs.google.com/presentation/d/1fKcTKwcglQ3bIJjBGuAvnpqK6TOEtZwuIJ5YHXuIK0w/edit?usp=sharing

Anna: In line annotations help to enrich data
... It is uses to measure how text changes over time

<tantek> Randall - thanks for posting it on the web! Hoping you can convert it to simple HTML too at some point :)

<Randall> yeah, totally

Anna: Annotation are not data you throw away, they grow with the text

<tantek> so you know, I could link to specific slides :)

<Randall> I was going to write it html-first, but I honestly just procrastinated too much and needed to bang it out quickly. presentation editor ftw.

<tantek> true

Anna: Annotations should have their own category

<tantek> I wonder if there is a Google Presentation exporter that produces decent simple HTML with URLs per slides

<KevinMarks> they used to (original format was s5) but I think they dump out svg now

<tantek> KevinMarks - that's too bad

Anna: You should be -citable -precise -annotation of segments/rgeion - Annotation extensible list of media types

<KevinMarks> pdf; pptx; svg; png; txt

<astearns> tantek: you might be able to export to PPT and from there to HTML :)

<KevinMarks> spot whats missing

Anna: Scholarly annotation requirements - Migrate -Copyright

<KevinMarks> the svg is also weird as they turn the text into paths

<tantek> astearns, sounds dicey

<Randall> tantek: it appears the presentation view does shave fragment identifiers that work to reference the pages

<KevinMarks> slideshare would mung the ppt to html

Anna: Storing bits of the annotations

<tantek> Randall - ah, cool, at least there's that!

Anna: You have to maintain the rights to the text
... Implementation concers
... -Our systems use Annotea, OA -shared generic backend, custom clients
... Discoverablitiy and sharability

Question: Paolo

scribe: "Do you have a feeling that a lot of the problems can be moved from the back end to the front end?"

<azaroth> Q from PaoloC: Can we move some of the complexity to the backend to make clients simpler?

Anna: Yes!
... We need to focus of more of these probibility questions to make improvements

<KevinMarks> Randall slid.es is HTML based (it's a UI for reveal.js)

<Randall> cool, thx

Speaker: Sean

Sean: About Logos - Digital library platform for bilbical studies
... - 40k+ books
... Editoral Annotation - Linguistic analysis - Alignment across language - Curated reference data annotated on on sourec data
... Why would a person buy from us rather that Kindle? Because we do a lot of annotations
... Ex: We convert languages for readers
... We allower user annotation, such as highlighting
... Using Annotation Documents - Display through visual fliters
... Areas of interest for open annotation - Standardized reference schemes - Bibliographic references
... We do a lot of word crossing

<KevinMarks> my notes are up at http://www.kevinmarks.com/w3cannotation.html

General Requirements on Annotation Models and Systems

Existing Annotation Systems & Implementation Issues

Speaker: Nick

Nick: I work on annoator
... We about to embark on something that could change the webpage as we know it
... Advanced data
... No one could predict that html would advance to what we are currently seeing on the web
... Stuff that we already do is annotating
... A couple people in the room have filled us in on this success
... Get gets way more extreme than that
... Such as deep zoom imagery
... Three things we will talk about - User interaction - Protocols - Data models
... An unlined wall are data models

One on the design pronicles is there should be one shared model

scribe: If we don't have that we will have extreme usability problems
... It is more ideal that one protocol will be user for alll
... The problems is to do annotation on the web is there are really only two ways to annotate
... A bookmark clips are soon going to die
... They will mostly likely be outdated in a few years
... They are not standardized
... Does anyone know of proposals to allow user trusted code to run?
... no!
... We specify everything
... We need to build pluralism

Speaker: James
... I come from a publishing background
... When I first started this process I asked colleagues what improvements they would like to see in annotations
... This is how we started making changes

<Randall> tantek: this is why I believe in webmention: "<bupkes> I'm very much a non-coder, non-technical person but like the idea of indieweb so am tinkering with my site. Have webmentions etc up and running through a bunch of WP plugins and stuff"

Speaker: We are going to talk about the authoring process

<tantek> Randall - yes we made major non-coder progress at IndieWebCampSF about a month ago. We've got non-coders doing POSSE and backfeed.

<Randall> that's soo sweet

Speaker: Annotations are currently limited
... Currenly it doesn't have a lot of annotation functionality
... Articles don't link the errors

<tantek> Randall - it's a direct result of: 1) as simple as possible standards -> 2) more implementations by more devs faster -> 3) faster usable iterations on those implementations -> 4) sooner emergence of non-coder usable software/plugins/tools.

Speaker: We have third party that track articles for referencing purposes
... Currently screen readers only have a contact list selection
... Orginally we started out with tear sheets
... We then moved to Word
... From there we moved to editior

<nickstenn> my slides: https://speakerdeck.com/nickstenning/annotation-user-interfaces-design-for-plurality or (persistent URL) https://whiteink.com/2014/04/w3-annotation-workshop/annotation-ui-plurality.pdf

Speaker: Currently working on simultaneous use
... We want multiple users to work from one page at the same time
... There is currently a way to tag users using annotations
... We hope our on going changes will encourage scholarly discussion
... Last three ares are - Curation - Storage -Out of Print

<azaroth_> Robert Casties: Q: Are you using / looking at any standards?

<azaroth_> Ana: Not yet, but we want standards to emerge to refer to

Speaker: Frederick

Frederick: We are working with educational groups in Finland
... The students are working with back end systems

<KevinMarks> pdf--

Frederick: Three use cases 1) students want to make notes=annotations
... 2) where you get assigments
... Foe example a work sheet
... You can annotated the question
... 3) Interaction betwwen the student and teacher
... The ability to use annotations is powerful
... You may want to know who did what
... Identity and access control
... Epub is evolving rapidly
... The education systems may move to online educational platforms

<nickstenn> KevinMarks: sorry... but again, it's about whether I want to spend half a day making an HTML and CSS based presentation authoring system work or whether I just want to make a presentation

Doug: Does anyone have requirement?

Phil: Bring in fragments into the learning management system

<KevinMarks> right. Though you can just write straight html into slides reveal.js, which I find pretty quick

Phil: There needs to be a way to ties two different sources

Alex: How to you measure where are annt ends and begins?
... This should be a requirement that we discuss

Anna: Break the categories

<nickstenn> KevinMarks: I did a presentation with reveal.js about 8 months ago and sorting layout took me about an order of magnitude longer than using Keynote. In an ideal world, where I had loads of time to write my presentation, then no doubt I would do this. Sadly, I don't live in that world :)

Anna: Elevaluate Protocols and user space

<KevinMarks> fair. Just make sure you export your keynotes so they don't rot

Rick: It's important to deal with levels of authority
... I need a way to apply authority to the inforamtion

Ivan: Somehow to report providence

Paolo: Lots of levels of prominence
... The levels can get over whelming

Philippe: We should be very careful about the views of certain annotations

<nickstenn> if anyone wants to chip in trying to capture the broad areas that people raise for later discussion, I've just started https://hackpad.com/Web-Annotation-areas-for-specification-Y5AkaI4LfAm

<nickstenn> and will export this at the end of the day if it becomes useful

Robust Anchoring

Tim: I'm going to talking about requirements in terms of robust anchoring
... Precision is important
... Recognize that annotations change
... One new kind of anchoring is a search functionality
... Anchoring methods should be fine grained for text
... Manually transcriptions of a wide range of work.
... Preserving providence

<KevinMarks> it's annotation turtles all the way down

Tim: Currently there is correlations between the pdf and the line you are correcting
... I would be nice to center the errors into one place for the user
... Working of the process of annotations

Existing Annotation Systems & Implementation Issues

Éric Focusing on Egyptology

scribe: We have different languages that we have combined on our webpage
... We published primarily in Epub
... No DRM
... you can double tab and zoom
... maps, when you click on the wall you can see up to 1000 mega pixels
... You can can specific parts of large images
... Something that is important in user usability
... The user needs to be able to easily reference information
... We transform paper books into ePub information
... The size of the text is adjusted
... Creating shortcuts
... We need all of the requirements of an annotaion anchor

Kevin: Best citation?

Frederick: Annotation link?

Eric: I want this person to be able to access all information easily

Fred: Kindle covers up sentences with annoting
... There is a problem when multiple people try to make highlights on one page
... The text selector is over loaded
... It's complicated and confusing
... Problems to solve this would be to create context, then anchoring it to the document.
... More space to compose
... Allow the reader space to anchor and reachor the annotations.
... Icons in the margin might help
... Mini paragraphs might help
... Define styles in CSS Print style. Print out annotation with the main text

Anna: Thank you for highlighting the experience between the user and the professor

Kristof: One of the things we change is usability formats
... For example PDF
... Changing the interface to work with Google doc
... Now we have a different format for multiple selectors
... We have a bunch of strategies to find the best match of text
... If everything fails to rely on fuzzy
... Two phase anchoring
... Problem: lazy rendering
... Proposed solution: two phrased anchoring

<nickstenn> more specifically, csillag is drawing a distinction between a) anchor dereferencing, and b) rendering highlights (say)

<KevinMarks> commenting on a removed item can be important

Kristof: Problem: Dymanic sites Proposed solution: dymanic anchoring

Kevin: How do you deal with unwanted annotation?

<nickstenn> Kristof: question was: removing annotations on removed content can be problematic, no?

Kevin: We are working it


Anna: How do you deal with many sets of annotations

<KevinMarks> Kevin: if you remove annotations when the document referenced is edited can't unfavourable ones be removed?

<KevinMarks> Kristof: yes, but we can keep orphaned annotations and possibly keep old versions

Anna: having page numbers of text

Speaker: Alex
... The current problem is when we transfer interfaces we lose information. We need to improve HTML

Doug: The working group should work of annotations for the web

Greg: Focus on establishing what a data model is

Doug: I'm talking specially about robust anchoring
... Having a central place where all data lays

Nick: .. When speaking about annotation content, there are some really diffuclt thing when communicating to the user.

Michael: I'm a user, how would I figure of annotations without compromising my privacy?

Randal: It might make sense to create hubs for privavcy

Michael: I still have a problem with monitors and anchors

Tantek: There might be a problem to the solution by hosting the anchor

… at annotation providers

… so how about the scenario of zero annotation provider?

… where the user hosts their own annotation on their own site

… hosts the anchor

… and hosts the context of the anchor as well

<azaroth> +1 to Tantek. Sharing can be with yourself, just across platforms/devices.

… with zero communication with other sites?

… that's a scenario that I'm interested in exploring.

Mitar: Creating annotations like Wiki

<Randall> tantek: +1

Speaker: ?

<nickstenn> Philippe

<nickstenn> Philippe != Phil Desenne

Phil: Large parts of the text could disappear
... How do you solve fuzzy anchoring?

<KevinMarks> how do you solve fuzzy anchoring for video

Speaker: Micheal

<nickstenn> I do hope we don't end up trying to standardise all these things

<nickstenn> otherwise vendors will take one look at the spec and burst out laughing :)

<KevinMarks> the fuzziness algortihms?

Speaker: We are trying to specify the ties

<KevinMarks> standardising how to store anchoring extracts seems a sweet spot

<KevinMarks> then they can all compete on fuzzy matching

<nickstenn> right, exactly

Dout: URL syntax allows you to specify a timestamp
... Gives a viewport

<nickstenn> having an API to discover and retrieve OA annotations for a resource is reasonable

<nickstenn> and for finding the possible anchors that could apply to the current repr

<KevinMarks> seems like multiple anchoring types that are aligned make sense to store

<KevinMarks> the twitter habit of citing >140 chars by using an image of the text is an example

Michael: How do we distinguish two documents?

<nickstenn> but standardising all possibly fuzzy anchoring algos on all possible media types...

<nickstenn> speaker last-but-one was Michael Scharf

Phil: There should be an image analysis

Ivan: We could be careful not to lock ourselves in our own community.
... Our documents should be shared throughout many communities.

<robcast> this is nick speaking

<tantek> twinkle = like = jazz hands

<tantek> twinkle down = dislike = squid hands

<astearns> we are now standardizing real-time physical annotations on the annotation discussion

<tantek> point at the person with index finger - I have a direct point on what is being said

<tantek> raise flat hand = q+

<tantek> hand in a cupped C = need clarification on current point

<tantek> KevinMarks

<azaroth> More information = more robustness

<Randall> yes yes yes

<Randall> also why URI fragment doesn't scale

<Randall> a target is an arbitrary complex bundle of resources (optionally w selectors)

<tantek> Speaker: Rob Sanderson

Kevin: Finding ways to link things for easy usability

Anna: When we talk about anchoring, focus on the content and not just the document level.

Mitar: We need to think about specification.

<KevinMarks> http://www.kevinmarks.com/w3cannotation.html updated with mroe notes

Data Annotation

<tantek> is there an archive of this channel?

<tantek> preferably with citable anchors for each statement :)

<ivan> scribenick: nickstenn

Raquel Alegre works on CHARMe, a European Commission funded project

scribe: climate data
... annotating climate-related datasets
... radar maps, etc
... advanced use cases, including:
... research on events timing
... focus on specific areas of the world
... intercompare datasets
... climate data comes in 2D, 3D, and 4D formats
... how to refer to a specific subset of these?
... OA SVG selectors are not enough
... dimensions of (e.g.) latitude, longitude, altitude, time
... looked into adding geographic selectors for these cases to OA
... CHARMe is just about climate data, but geographical selectors have much broader use than this
... important that schemas used for describing annotation anchoring have meaning outside of web pages

Tim Cole: there's a difference between say, a sensor reading "zero volts" and how that translates into, say, a temperature of "54 fahrenheit"

scribe: so keeping raw data rather than just interpretations of that topic is important

Doug: discussion on mailing list about being able to argue about sources of data
... e.g. "that sensor has a known bug in that range" etc

Storage and APIs

Gregg: Hydra is an outgrowth of work on JSON-LD
... JSON-LD goes part of the way towards expressing data in an unambiguous way by describing data in terms of well-known schemas
... what Hydra does is provide a vocabulary for describing operations on and relationships between entities in these APIs
... Problem: how do I use an API without specifically coding for it?
... Answer: use vocabulary to define operations on classes and properties
... Proposition: annotations are the result of operations on entities or the relationships between entities
... Those results are also entities, which can also be operated upon


scribe: Use case: given a profile page for a sports personality, how can I interact with that entity -- (follow, {dis,}like, share, question, suggest relation)

Jason: From IEEE Learning Technology, talking about the Experience API (xAPI)
... Project "Tin Can API"
... based on http://activitystrea.ms
... description of activities of form <actor> <verb> <object>
... originally developed for learning management, course software, etc.
... now branching out: simulators, virtual worlds, group-based learning, etc.
... xAPI is both machine- and human-readable
... xAPI specs: https://github.com/adlnet/xAPI-Spec/
... pretty good adoption already, both open-source and commercial
... IEEE Actionable Data Book (ADB)
... EPUB3 + xAPI in the iBooks, Reasium, EPUB.js, Calibre
... additional areas to investigate: widgets, client/reader xAPI implementation, annotator, ...
... [Jason is demonstrating some of the integrations]
... interested in opportunities to collaborate between OA and xAPI

KarimKhan: annotations on ePub could be very valuable in annotating research

FrederickHirsch: question -- how long-term is "long-term stable" in the context of ePubs

TydeRichards: one motivation comes from the aviation industry where training docs need to last on the order of 30 years

JakeHartnell: Open Annotation Architecture and Scope
... I love imagining the future -- let's imagine private annotation in browsers in 2018
... Web Annotation is successfully implemented!
... so here we are on a web page

<KevinMarks> anyone else reminded of SideWiki?

JakeHartnell: we open up an annotation editor --- an HTML document editor
... what happens when you make an annotation?

<Randall> KevinMarks: always

<shepazu> KevinMarks, nope, not yet

<shepazu> :P

JakeHartnell: browser captures selection information which it passes to a new html doc
... channels: aka annotation doc stores
... each channel can have a variety of settings associated
... users control which annotation channels they subscribe to

<KevinMarks> channels aka hashtags

JakeHartnell: annotations have many use cases

<Randall> KevinMarks: i'm not sure about that

<Randall> that's a pretty centralized version of channels

JakeHartnell: [Jake is showing a mockup of the chrome settings screen of the future, with channels for hypothes.is, microsoft office, newspapers, etc.]

<KevinMarks> how are hashtags centralised?

JakeHartnell: the browser queries all the channels a user is listening on
... a bit like RSS
... Anno docs are loaded into sidebar where they can be seen
... we can filter them,
... we want to be able to (as users) switch between social contexts

<Randall> KevinMarks: well, I was presuming that #term was enough to know the URI for subscribing

<Randall> if that's not what you meant, then I'm wrong

<KevinMarks> you can subscribe to it on services you use, but it maps across services

JakeHartnell: multiple levels of annoations
... 1. annotation as advanced linking
... 2. user control --- give people the greatest amount of control in managing what they see
... important for combating noise
... 3. a space --- browser should provide a space for attached documents to live and be viewed, and should be a store for personal notes, documents, etc.
... We need people to be able to link their thoughts to things.

<KevinMarks> [just one browser? I have at least 5 open at the moment]

TimCole: status quo is that annotation stores control where those annotations appear
... you've talked about user control
... I think there needs to be a half-way point
... I want to know where annotations are being displayed

BarbaraTien: I liked the future world view aspect
... problem of trolls and so on
... but imagine a politically charged scenario: imagine a fox news filter on a whitehouse channel

JakeHartnell: people will always choose the moderation policies they want
... people will have different models of adding secondary content
... provide tools that allow for variety
... comes back to providing space

PhilippeAigrain: the user must have sovereign control over their space

<Randall> Jake: Where can we actually put this stuff in the browser?

<Randall> ... as we grow, from bookmarklets and extensions, ...

<Randall> ... how can we allow a model where we allow external services into the page

Genesis: how do we encourage participation? bring down the barrier to entry
... installing a new app, learning a new tool is difficult

Jake: when you give developers the tools to do these things, people *will* go out and build things that resonate with all kinds of groups (cites example of RapGenius)

Doug: if we can get annotations in the browser, then that hugely improves discovery

KarimKhan: analogy with "share your location with the webpage?" toolbar -- "do you want to join the discussion on this page?"

Mitar: multiple languages problem: how do we annotate documents which are translated into multiple pages

PhilippeAigrain: there's a tool called TLhub -- social network of translation -- interesting concepts which could be appropriated

Doug: accept the use-case, but we need to focus on what we can conceivably standardise in the next 2 years

GregKellogg: what we do could either enable data silos, or we can try to ensure interoperability and build something shared and distributed

FrederickHirsch: are we going to do the charter at the end of the day, or should we be trying to narrow down as we go

Doug: yes and yes

<tantek> <br>

Accessibility & legal issues

GerardoCapiel: going to be talking about how annotation can be a powerful tool for discussions of non-textual content
... [video of BookShare ... reading books using a screenreader + braille display]
... [showing an example of an image without any alt text: useless in a screenreader]
... images tend to lack descriptions
... mathematics tends to be rendered to images rather than using MathML
... video description has even less support
... [demo of youdescribe, in which a video pauses occasionally to play an audio description of video content]
... state of the world: blind and visually impaired (VI) students often need help navigating content
... disparate enabling tools
... [showing Poet Image Description: navigate all images in document and annotate them]
... [showing WebVisum ... a tool to allow sighted users to annotate visual content for blind and VI people]
... conclusion: ann powerful tool for addressing accessibility of non-textual content on the web and in eBooks
... it's happening today with disparate tools
... we need unified and standard approaches

discussion about how to select text with screen readers ... it can be done

KarimKhan: what does MathML offer over images for mathematics?

GerardoCapiel: it's a machine-readable format
... VoiceOver knows how to read MathML
... as do a number of other tools
... it's important to be able to navigate the mathematics with the screenreader

TimCole: lots of kinds of mathml.
... also latex can be translated to speech

GerardoCapiel: Presentation MathML mostly used. Content MathML barely adopted

PuneetKishor: First of all IANAL
... my intent is to avoid unleashing rats [from the rat's nest of legal and copyright issues]
... i work at Creative Commons
... my mission in life is to ensure that things remain as open and free as possible, unfettered by the law
... if the annotation system does become rich and thriving, then the issue of intellectual property could become a rat's nest
... we want to avoid that
... i have neither slides nor solutions
... for the people who are designing these systems: consider building the mechanisms that make clear the legal status of annotations *from the design stage*
... and not as an afterthought
... it's not clear whether annotations will have copyright implications or not
... they may not have enough creative content in them to be copyrightable
... but perhaps over time they will
... three things I think of [on this topic]:
... 1. every annotation should carry its legal status with it. its licence is metadata.
... annotations can build on other annotations, so it can be important to track provenance of annotations [and annotation chains]
... problems: how do you express the licence?
... binding: do you agree to the licence by simply visiting and view the annotation?

Ivan: do you envisage that the current CC licence terms are ok [for this context]?

PuneetKishor: I do. Previous iterations of CC licences perhaps not, but most problems around use of CC licences for data have been addressed in the latest version (4.0)

<gcapiel> here's a copy of my presentation: http://www.slideshare.net/gcapiel/i-annotate-prez-2014

PuneetKishor: hopefully they won't need updating for a decade

GregKellogg: the copyright implications of linking to copyrighted documents have been discussed at length
... do you think there will be similar issues around "right to annotate" as with "right to link"

PuneetKishor: I don't really know. Possibility that Terms of Service could restrict those rights.

Doug: Important to draw the distinction between licence of source document and licence of actual annotations

PuneetKishor: yes, I'm just talking about licences of annotations. licence of underlying document will be set by the document author.

TimCole: how do we deal with the assertion of authorship on annotations?

PuneetKishor: I don't really know. CC only really cares about IP, not assertions of authorship, etc.

AnnaGerber: be careful around distinction between licence of annotation and licence of annotation bodies
... specifically, the creator of the annotation may not be the author of one (or more) of the annotation bodies

Doug: [clarifies terms from OA model]

PuneetKishor: you can only licence what you create

Doug: you might have licence per-body as well as per-annotation

PuneetKishor: CC is not just about legal obligations and restrictions, but also about establishing norms about behaviour
... attaching a licence to content doesn't necessarily mean that you have the legal right to make those assertions
... but it might still serve to state the social norms

<KevinMarks> everything is copyright by default

<KevinMarks> CC lets you disclaim parts of this

FrederickHirsch: why do annotations require copyright?

PuneetKishor: I didn't say they did. They might, they might not. But if they do, you should build support for describing legal status from the beginning.

<scribe> scribe: RandallLeeds

<ivan> scribenick: Randall

Tantek: On copyright, I've found that both APA and MLA style for tweets include the entire tweet and say nothing about licensing
... if there are multiple, respected organizations encouraging this, it's an assertion about de facto rights of small content
... On copyright, I've found that both APA and MLA style for tweets include the entire tweet and say nothing about licensing
... if there are multiple, respected organizations encouraging this, it's an assertion about de facto rights of small content

PuneetKishor: an annotation could be very long

<tantek> APA citation style of tweets and research: http://microformats.org/wiki/citation-formats#APA_permalink

PuneetKishor: I make a massive, 1000 word annotation, on a piece of poetry, that would implicate copyright by most standards

<tantek> MLA citation style of tweets and research: http://microformats.org/wiki/citation-formats#MLA_note_permalink

<KevinMarks> Twitter's official tweet citation spec also includes the full text of the tweet

ivan: whatever system is developed at w3c should have the necessary hooks for copyright info

<KevinMarks> which provides a fallback if the tweet is deleted or twitter is down, or you're in Turkey

PuneetKishor: we (CC) have already created the license. your job is to figure out where to slot it.

Charter discussions

Ivan: [showing the draft charter: http://www.w3.org/2014/01/Ann-charter.html]

<tantek> Please share URLs for your slides.

<tantek> So we can have a canonical URL to annotate with comments ;)

<tantek> (per slide if possible, I.e. HTML presentations with URL per slide)

Ivan: I hope it will help us drive this discussion
... The really essential parts of the charter are the six bullet points that categorize the distinct things this working group can tackle
... Abstract data model (based on Open Annotation work). We can discuss whether/how to fast track this.
... Vocabulary is very related to data model
... The annotations have to be serialized. For those who understand RDF "no big deal" but things are not that simple because quite a lot of people for whom it is too complicated.
... We may want to explore adding to HTML with microdata or new elements to provide a different expression
... The annotations have to be stored. So some discussion of serialization
... They also need some APIs

<tantek> ( /me humbly contributes http://microformats.org/wiki/h-cite as a simple vocabulary for web citations )

Ivan: Robust anchoring.
... Client-side API. Has to be specified because we want a portable layer on that platform.
... Nothing in here is related to user interface. It cannot/should not be standardized at the w3c. It comes into the charter maybe only as "client-side API" and needs to be driven by use cases.
... The way we envision the organization of the WG is essentially two task forces: data task force and client task force
... Of course they are not disjoint
... These are not the only documents that we can produce. There can be other documents that are not recommendation track.
... We might have requirements documents for annotation as a whole in different environments

DougSchepers: Robust anchoring is not only a problem for annotations
... If we can we want to make generalized models
... We tried to break these sections down into re-usable parts
... Maybe the WG doesn't address everything. CSS WG might address styling, for instance.
... and decide if there are large problems to discuss for charter or not

Frederick: How do other reading environments have to be addressed in this WG

Ivan: an epub reader is the same as a browser, to me. I don't want to make a difference here.
... This may be the first time we have an inverted constituency: publishers around the same table as the traditional browsers

DougSchepers: The basic issue of how we get interop is by testing and getting the right players (browsers and readers)
... if we don't have that... the value is in question
... a lot of people feel the data model is pretty mature. we might get interop on some things but not all.
... the whole goal of the WG is to establish this. we don't have to resolve it here.

RobSanderson: As much as I like loose coupling, the vocab/serialization/data model will be hard to tease apart.

Ivan: that's absolutely correct
... OTOH it's true that the OA data model is already used out side the www
... our goal here is to concentrate on web issues, but some things (if properly sold) would be used by, e.g., the library community

<fjh> would like the charter to clarify groupings of 1-6 that need to be interop tested in conjunction and expectations for number of independent implementations and types

PaoloCiccarese: We talked about provenance and copyright. We know the current model can accommodate.

<fjh> also might want to clarify expectations around adoption

PaoloCiccarese: What do we do with other media, like video and images, that don't seem to be address in the charter

DougSchepers: I'd like to get a sense of the room quickly
... I'd like to get a show of happiness if you think we should explicitly mention provenance and licensing in the charter
... Who thinks provenance should be mentioned?

<nickstenn> RESOLUTION: the group generally agrees that provenance of annotations should be included in the charter

RESOLUTION: the group is generally agreed that provenance was an important thing to mention

DougSchepers: scoping is about what we should talk about but also it might have implications for who shows up to the table

Ivan: Because of existing IP issues, if the WG decides that there is another item to standardize, formally we the group must be rechartered
... it's important we have here everything we want to do
... but that doesn't mean everything has to happen

NickStenning: The legal status of annotations and annotation bodies might be themes but might not necessarily stand on their own as a top level item on the charter

DougSchepers: To be clear, to put something about provenance in one of the bullet points, but not the primary description

RobSanderson: further clarification: just because we don't specify something doesn't mean we can't do it if it falls under some sesction we did call out
... this is to advertise the group to people who might want to get involved

DougSchepers: this is more art than science. it's like refrigerator art.

KarimKhan: It's important to allow people to put source information

PhilippeAigrain: Small comment on licensing. There are some countries where rights can be restricted based on contractual obligations.
... the way licensing is mentioned, much of the rights of use of annotations...

Ivan: what we propose is only to add a sentence like "the annotation model should make it possible to add provenance and copyright information." full stop.
... that's all we would do.

Mitar: I agree. But it should not be written that provenance is necessary linked only to the model.

DougSchepers: Ivan gave a great example.
... Three things from Paolo
... should we talk about licensing, in addition to provenance
... who thinks we should? [some hands]
... who thinks we should not? [very few]

<nickstenn> RESOLUTION: the charter may talk about licensing

DougSchepers: We're just trying to get broad ideas here.

GregKellogg: pertinent to Ivan's characterization of provenance, the data model should allow the expression of licensing information.

<Mitar> Mitar: But it should not be written that provenance is necessary linked only to the model.

NickStenning: Thank you, Ivan, for clarifying the need to re-charter if things aren't specified up front
... but with that in mind, I wonder whether the link anchoring thing, which talks about algorithms rather than the process, might be a deep hole
... trying to specify algorithms might be extremely difficult and burn many cycles

DougSchepers: I'm going to interpret that
... we want to go through each of these in depth
... But first, should we do other media types?

TimCole: some are important, some less so [? something like this]

DougSchepers: whatever mechanism we have, it should be extensible
... we aren't going to do everything in a year or two years of work
... Show of hands. Assuming we have something about robust anchoring, do we agree there should be an extensible mechanism for other media types?

RESOLUTION: if we do have robust anchoring in the charter, it should be extensible for different media types

FrederickHirsch: Just to clarify, we want the data model obviously to annotated things like maps etc. Do we really need to spell it out?

DougSchepers: Example of why it's important
... if someone has a patent specifically on linking to a map
... they're going to want to know if that's in scope

MichaelScharf: in the 6 points, do we include the injection of annotations?

DougSchepers: We're not quite there yet... your comments can ncome back on the stack

PaoloCiccarese: We're thinking about all these different communities. If we keep the extensible charter language we might be able to use the community group as an incubator.

DougSchepers: Show of hands if there's something missing? I see some confused faces.

Speaker: I was curious about accessibility. Where does this fall in here? Does it? If not, why? (Things like ARIA)

DougSchepers: We maybe should have said something about accessibility, but it's a requirement for W3C
... does everyone agree that we should address accessibility [broad concensus]

Ivan: The charter has liasons, particular people who link this work to other working groups that we can explicitly ask for comment
... we have to and we want to connect to those groups to look over our shoulder

<nickstenn> just throwing this in here as it seems appropriate about now: http://imgur.com/yuNGOgz

DougSchepers: Accessibility is implicit in all of them

Mitar: Privacy and security

DougSchepers: Yes. These are what we call "horizontals"

MichaelScharf: We are doing in some cases injection of 3rd party code. Should we spell that out in the charter?
... This specific problem of cross page scripting.

NickStenning: To rephrase I think the question is that specifying all possible UIs is out of scope.

DougSchepers: or any at all

NickStenning: But it needs to be plausible that people can create them in a standards-based way
... at the moment I posit is that it is not
... because browser extensions are vendor specific
... but I do not think it's the role of this group
... though it may very well be that if we don't have some standard way of building and shipping these components all will be for naught
... is there another WG we should have as liason for this?

DougSchepers: It's too big a question for us to solve.
... Maybe browsers will see the light and do it themselves.

KarimKhan: As long as we have a way to manage this place and moderate this spillover there is no problem.

DougSchepers: Should this group's charter include a mechanism to enable extension-like capabilities?

Mitar: maybe we can add something smaller -- analyze the scoping of that?

DougSchepers: We shouldn't do that. That's a whole other group.
... Should the charter include mechanisms to offer privileged access to webpages for extension code?

NickStenning: to me that sounds like it should be a liason with webappsec

TimCole: how do you do number 6 [client APIs]?

DougSchepers: You do [number 6]

<Mitar> *but we could then add exploring the problem space to the charter?*

NickStenning: It's helpful to say to people why the answer might be no.
... Such as having it happen through liason

Tantek: one of the ways of saying no should be saying "we want to delegate this to another group"
... it doesn't seem reasonable to say "we only want to talk about deliverables"

DougSchepers: First I want to resolve the question of do we think this belongs in the charter?
... Up or down vote twinkles?

RESOLUTION: we will not explicitly put in scope privileged access to third-party web pages' execution context

GregKellogg: I want to clarify that if, a year down the road, the group can now not do it because it's not chartered?

DougSchepers: Correct

Ivan: Correct

DougSchepers: The group will be rechartered in 2 years anyway

Mitar: we cannot specify what methods there should be for injection, but I think we should specify requirements for annotations to work

DougSchepers: I respectfully disagree

<tantek> I'll ask in IRC: I'd like every mention of "Microdata or RDFa" in the charter to explicitly include microformats as well, e.g. "Microdata, RDFa, or microformats" e.g. in the charter it says "(decorated with Microdata or RDFa)"

I think that should be in a community group

scribe: I think that should be in a community group

DougSchepers: there's nothing stopping us from talking about other topics with other community groups

Mitar: what about federation and decentralization?

DougSchepers: I would like to see it mentioned in the specification

KarimKhan: If you have means to use local storage and local control offloads the problem of that
... if the API says someone is going to be responsible for storage

DougSchepers: process [we'll come back as we go through the points invidually]

Tantek: I would say we need to consider going against the previous framing of excluding UI and instead include discussion/research/documentation (if not specification) so that we can do so in the WG and document it publicly

DougSchepers: I love to have discussions about UI. Browser vendors go crazy and hate us when we do it.

Tantek: Not specification.

DougSchepers: I think it could be within the scope. Or we could form a community group.

Ivan: I would actually be in favor to put in the charter some thing that says the group will produce a note on some UI ideas or whatever
... it may well be that during the charter discussion the browser vendors will say 'no way'
... that's why they are members

RobSanderson: For number 6, we could say that we will create a requirements document

Ivan: I would say requirements is even too strong

Tantek: "non-normative examples"
... I want the group to have the freedom within the scope to have those discussions

DougSchepers: Who thinks we should mention "non-normative" (not requirements, just exploration) of UI
... I see roughly equal numbers of should and shuold not

AnnaGerber: I just had a point about scope. We already have a lot of work for 2 years.
... UI is a giant rabbit hole

<tantek> I agree that UI is a giant rabbit hole, and yet, if you don't have UI examples driving your API/protocols/formats requirements, then you're going to have a bad design.

<tantek> But I'll submit to the rest of the group on this. Don't have strong feelings.

NickStenning: closely related is whether there should be discussions about what is required for standardised extensible UIs

DougSchepers: I think we should move on to discuss the individual items

Ivan: we added already the approach to base the data model on open annotation data model
... I don't know if this is something we want to discuss or if it's just sort of agreed

FrederickHirsch: I was suggesting we try to move it to rec quickly and say that in the charter
... there's a lot of work in this charter and that model seems well thought through

DougSchepers: I don't see a problem with a v1 and a v1.1 quickly
... unless things will drastically change

Ivan: I don't see a real value of rubber stamping a 1.0 version with the knowledge that we'll work again on it
... because the current CG document is already out and stable
... putting a stamp on it just to revise it ... I don't see a reason for that

TimCole: The problem with the current model is the need for an HTML serialization

DougSchepers: An explicit HTML serialization - I don't see - I think that just because something is stable doesn't mean there isn't value in the WG stamp
... Is there anybody here not comfortable with the language in the charter about OA?
... [no]
... Do we need to specify higher than an annotation in the data model

PaoloCiccarese: If we don't get there we will have many solutions

AnnaGerber: There are already specifications for collecting groups of resources. We should focus on annotation.

Ivan: On one hand, we say that the data model is based on Open Annotation but we have to look at it again
... On the other hand, we say we have to have a use case document

<tantek> How is there going to be a use-case document without discussions and examples of user interfaces?

<tantek> And user interactions?

Ivan: I don't see why I would pull out, explicitly, this feature and put this in the charter

DougSchepers: We'll consider adding sets of annotations, but we're not going to [mumble mumble mumble]

<fjh> i suggest we also consider moving model to Rec as 1.0 then consider 1.1 as needed to consider milestones and timeline

Philippe: the first three items would be compulsory in any discussion of this type

<Mitar> *are we having comments on all points or just on the first one?"

Philippe: the next three points - for example, there are two types of APIs mentioned - this is something that we do need lots of exploration
... from practical experience it's not easy to understand what feature fits in each

<ld> Philippe I think?

DougSchepers: I think the first 3 things are very closely related
... Does anyone feel like any of the first three should be dropped?
... [no one]
... That leaves us at the RESTful API
... Maybe we just reference something else or develop something new and magical
... Question on the table is should it be there at all

FrederickHirsch: I think search needs to be in this list
... for serialization JSON should be a requirement of the specification
... JSON-LD, actually

DougSchepers: We can leave requirement or not to the chairs, but search is non-controversal.

RESOLUTION: Discussion of APIs should also include search


Tantek: I thought you mentioned activitystreams, should it be mentioned explicitly?

DougSchepers: I'd rather not say how we're going to solve it in the charter

Tantek: I would advice not specifying activitystreams in the charter
... there's already a WG trying to get chartered for this
... and some organizations are objecting strongly to mentioning activitystreams

Mitar: I would propose we change RESTful API to just HTTP based API
... for instance, properly RESTful API maybe doesn't encompass push

RandallLeeds: Is "discovery" separate and worth mentioning with "search"

DougSchepers: I think we can mention a number of verbs and also leave it open

Ivan: agreed

DougSchepers: I think we should keep client-side API here, so we give ourselves scope even though it's ill defined
... objections?
... [none]

RESOLUTION: keep client-side APIs in the charter

<tantek> re: microdata or RDFa - alternatively, drop all mention of Microdata or RDFa, e.g. drop "(decorated with Microdata or RDFa)" from the charter, and leave it up to group discussions (rather than explicitly)

<Mitar> RESOLUTION: it is a friendly suggestion to change RESTful API to just HTTP based API

<tantek> Randall - I think that would be the simplest solution. And makes charter shorter :)

DougSchepers: How do people feel about getting less strong about explicitly specifying algorithms [for anchoring]? [some people]
... who feels it must be there? [crickets]
... who feels that we do need to address robust anchoring? [a gaggle]
... we can have a conversation later in the day

TedO'Connor: it's not clear to me that this shouldn't be in the charter at all

scribe: this group can define various sorts of requirements for platform features to enable that
... but I think we should stop short of saying it explicitly

DougSchepers: This group may not make all the deliverables.
... I'm not certain how we could put it in scope and not work on it

Ivan: In a sense, without this getting resolved, the whole thing might be useless (that's a bit extreme)
... but I think it should be reworded in a way that emphasizes the accessibility of it
... there should be a structure that works for the simple cases

TedO'Connor: If it's so important how could we leave it out? Maybe we should even have two groups?

Ivan: We discussed two groups. But it might create problems. Independent entities not sync'd with one another.

RobSanderson: If we don't call it out as a separate data point
... then it rolls up into abstract data model
... do we have to have a deliverable because it's so important to solve that problem?

DougSchepers: the only pragmatic reason to leave it out is IP to me
... since there is so much struggle, i believe it deserves to stand as a separate point
... two is the robustness aspect, which is worth defining to some degree
... it's clear that it would already be important to achieve the first part [which may collapse into abstract data model]
... but I'm not sure the expression "an algorithm" is the best one
... I would suggest making it implicitly part of the data model ties it too closely to annotation

MichaelScharf: There is the data model that is stored and there is the algorithm that finds the references
... the algorithms may change over time, but the data must be specified

RandallLeeds: I think maybe it's a good moment to try to place a resolution on removing/altering the explicit mention of "algorithm"?

DougSchepers: [something something something, essentially, "a mechanism"]

<nickstenn> so, what exists in point five

<nickstenn> but with s/An algorithm defining a set of heuristics/A mechanism/

<nickstenn> and s/an HTML5 DOM/a document/

<nickstenn> and s/$/with an extensible mechanism for alternate data types/

RESOLUTION: We will change robust link anchoring to something like: "A mechanism to determine a passage or portion of media that may serve as the target of an annotation with allowance for some degree of document changes and extensible media type support"

RickJohnson: If one of the outcomes of this is working together with epub than this has to be there

DougSchepers: How about, "including, but not limited to, HTML5?"
... I would like, at this point, to call this to a close, but first this last bit
... I'd like everybody to take a deep breath
... think, "are you happy about the outcome that a group like this might have"
... does this seem like something you'd like to participate in? please twinkle your fingers [so much twinkling]
... I think at this point we should retire to the bar.

Ivan: Before we go... I think we owe a thanks to Hypothes.is for organizing
... I think we are adjourned. So recording can stop.

oh, you did it


Summary of Action Items

[End of minutes]

Minutes formatted by David Booth's scribe.perl version 1.135 (CVS log)
$Date: 2014-04-03 17:38:43 $