W3C Video on the Web Workshop 12-13 December 2007, San Jose, California

Workshop Minutes

Attendees

Present
Adele Peterson (Apple, Inc., San Jose), Amit Sheth (Wright State University, San Jose), Andrew Fischer (Via Licensing Corporation, San Jose), Ansuman Tapan Satpathy (Motorola Inc., San Jose (first day only)), Antti Koivisto (Apple, Inc., San Jose), Bruce Fairman (Sony, San Jose (first day only)), Chris Double (Mozilla Corporation, San Jose), Chris Lilley (W3C, Brussels), Dan Connolly (W3C, San Jose), Dan Shanori (Accenture, San Jose), Davor Orlic (IJS, San Jose), Don Brutzman (Web3D Consortium, San Jose), Douglas Schepers (W3C, San Jose), Drew Major (Cisco Systems, San Jose), Elisabeth Freeman (Walt Disney Internet Group, San Jose), Emmy Huang (Adobe Systems, Inc., San Jose), Eric Carlson (Apple, Inc., San Jose), Eric Hyche (RealNetworks, San Jose), Erik Mannens (IBBT, Brussels), Franck Denoual (Canon, Inc., Brussels), Gerard Fernando (Sun Microsystems, Inc., San Jose), Glenn Goldstein (MTV Networks, San Jose), Guillaume Olivrin (Meraka Institute, San Jose), Ian Blaine (thePlatform, San Jose), Ian Burnett (University of Wollongong, San Jose (first day only)), Jack Jansen (Center for Mathematics and Computer Science (CWI), Brussels), Jason Gaedtke (CableLabs, San Jose), Jeff Dinkins (Sun Microsystems, Inc., San Jose (first day only)), Jeffrey Campbell (Cisco Systems, San Jose), Jennifer Taylor (Adobe Systems, Inc., San Jose), Jeremy Condon (Move Networks, San Jose), John Toebes (Cisco Systems, San Jose), Julie Lofton (Hot Potato, Inc., San Jose), Karen Myers (W3C, San Jose), Keith Lantz (Cisco Systems, San Jose), Kevin Calhoun (Apple, Inc., San Jose (first day only)), Kevin Lynch (Adobe Systems, Inc., San Jose), Larry Socher (Accenture, San Jose), Loren Larsen (Move Networks, San Jose), Lynda Hardman (Center for Mathematics and Computer Science (CWI), Brussels), Maarten Verwaest (VRT medialab, Brussels), Marc Owerfeldt (Sun Microsystems, Inc., San Jose (first day only)), Mark Bakies (Cisco Systems, San Jose), Mark Kortekaas (CBS Interactive, San Jose), Martine Roeleveld (Center for Mathematics and Computer Science (CWI), Brussels (first day only)), Mary-Luc Champel (Thomson, Brussels), Matthew Patulski (Capgemini, San Jose), Michael Dale (University Of California at Santa Cruz, San Jose), Michael Wise (Turner Broadcasting System, Inc., San Jose), Neil Izenberg (Nemours Foundation, San Jose (first day only)), Nimish Radia (Sun Microsystems, Inc., San Jose), Olivier Poitrey (Dailymotion, Brussels), Paul Bosco (Cisco Systems, San Jose), Peter Lambert (Ghent University - IBBT, Brussels), Philippe Le Hégaret (W3C, San Jose), Rajeev Mohindra (Motorola Inc., San Jose (first day only)), Rigo Wenning (W3C, Brussels), Rob Glidden (Sun Microsystems, Inc., San Jose), Robert Freund (Hitachi, Ltd., San Jose), Saurabh Mathur (Thomson, San Jose), Sebastjan Mislej (IJS, San Jose), Sharon Huang (Motorola Inc., San Jose (first day only)), Shashi Seth (YouTube, San Jose (first day only)), Silvia Pfeiffer (Annodex Association, San Jose), Soohong Daniel Park (Samsung Electronics Co., Ltd., San Jose), Stephan Wenger (Nokia Corporation, San Jose), Stephane Enten (Dailymotion, Brussels), Steve Bratt (W3C, San Jose), Steve McQuade (Walt Disney Internet Group, San Jose), Steven Xu (Motorola Inc., San Jose (first day only)), Thomas Roessler (W3C, Brussels), Tony Wyant (Sun Microsystems, Inc., San Jose)
In addition, several individuals were following the meeting using IRC, occasionally making comments in the channel. Most of IRC comments have not been retained here for several reasons but mainly clarity. Michael Hausenblas (Joanneum Research) and Raphael Troncy (CWI) are among the IRC commenters below but not listed as present above.
Chair
Ian Blaine, Paul Bosco, Philippe Le Hégaret
Scribe
Thomas Roessler, Karen Myers, Steve Bratt, Bob Freund, John Toebes, Don Brutzman, Dan Connolly

Contents

  1. Video 2.0: Transition from Broadcast to the Internet, Larry Socher, Dipan Patel (Accenture)
  2. Video - Any Device, Anytime, Anywhere, Ansuman Tapan Satpathy (Motorola)
  3. Strategic thinking for Web Video, Franck Denoual (Canon)
  4. Hitachi Position Paper: Video on the Web, Robert Freund (Hitachi)
  5. Sony Position Paper for the W3C Video on the Web Workshop, Bruce Fairman (Sony Electronics)
  6. Video on the Web, Kevin Lynch (Adobe Systems)
  7. Standardized multimedia elements in HTML5, Kevin Calhoun, Eric Carlson, Adele Peterson, Antti Koivisto (Apple)
  8. Thoughts on Future Requirements for Video on the Web, Eric Hyche (RealNetworks)
  9. Position Paper for the W3C Video On The Web Workshop, Chris Double (Mozilla Corporation)
  10. Web Architecture and Codec Considerations for Audio-Visual Services, Stephan Wenger (Nokia)
  11. Video on the Web at the Walt Disney Internet Group, Elisabeth Freeman, Steve McQuade (Walt Disney Internet Group)
  12. Potential for Internet Video Monetization, Shashi Seth (YouTube)
  13. Medical Uses of Video, Neil Izenberg, M.D. (Nemours Center for Children's Health Media / KidsHealth.org)
  14. Video as a first-class object for promoting scientific research, innovation and training, Sebastjan Mislej, Davor Orlic (Department of Knowledge Technologies and Center for Knowledge Transfer, Jozef Stefan Institute)
  15. Is Video on the Web for Sign Languages, Guillaume Jean-Louis Olivrin (Meraka Institute)
  16. Video Requirements for Web-based Virtual Environments using Extensible 3D (X3D) Graphics, Don Brutzman (Web3D Consortium, Naval Postgraduate School, Monterey California USA)
  17. Discussion
  18. Identifying Spatial and Temporal Media Fragments on the Web, Lynda Hardman (Centrum voor Wiskunde en Informatica (CWI)/Joanneum Research, Austria)
  19. Hyperlinking to time offsets: The temporal URI specification, Silvia Pfeiffer (Annodex Association)
  20. Video on the Web: Experiences from SMIL and from the Ambulant Annotator, A.J. Jansen (Centrum voor Wiskunde en Informatica (CWI))
  21. First Day wrap-up
  22. b2b+social media=p2p, Matthew Patulski (Cap Gemini)
  23. W3C Video on the Web Workshop, Jason Gaedtke (CableLabs)
  24. What would make the Web a true video platform, Loren Larsen (Move Networks)
  25. Methods for Preserving Audiovisual Metadata, Michael Wise (Turner Broadcasting System)
  26. Video Recommendation Based on User Preference on the Web, Soohong Daniel Park (Samsung Electronics)
  27. Standardizing the Syndication of Media with Advertising on the Web, Mark Kortekaas (CBS Interactive)
  28. Video on the Semantic Sensor Web, Amit Sheth (Wright State University)
  29. Discussion
  30. Enabling A Richer Video Experience With Metadata, John Toebes (Cisco Systems)
  31. Handing Over the Keys: User-Controlled Metadata, Doug Schepers (W3C)
  32. Metadata For User Generated Video, Drew Major (Cisco)
  33. Metavid & Free Online Video, Michael Dale (University Of California at Santa Cruz)
  34. Architecture of a Video Web - Experience with Annodex, Silvia Pfeiffer (Annodex Association)
  35. Future of Video and Next Steps Panel

Video 2.0: Transition from Broadcast to the Internet, Larry Socher, Dipan Patel (Accenture)

[Slides, Paper]

Larry Socher: Consumers want more control of their viewing
... move from pushing video to pulling it
... more control; different ways to see it
... few real-time things left
... most content could be made non-linear
... big problem is lack of the right content; libraries need to be built up
... desire fro user-generated content (YouTube). new devices. rip and go model
... iTunes, Zune,
... take PVR content along
... challenge: Digital Rights
... general belief that DRM will diminish over the years

<Dan Connolly> "Apple has sold more than 2B songs a 99 cents but made limited any margins"

Larry Socher: but there are some niches for it
... make sure content owners get paid
... but enable rip & go

<John> One take away I get from the people attempting to do this is that they aren't getting the content in a form that they want to consume it.

Larry Socher: mobile: Europe & Asia ahead

<John> hence doing the work to transform it.

Larry Socher: expect US to leapfrog market
... richer client experiences
... time & place shifting
... PVR changing TV
... fundamentally changes the way you watch it
... time & place shifting transitional effect
... will be around while digital rights are being sorted out ....
... Race to deliver 3 screen experience is on ...
... convergence is becoming a reality ..
... questions? ...

Video - Any Device, Anytime, Anywhere, Ansuman Tapan Satpathy (Motorola)

[Slides, Paper]

Ian Blaine: do you see some answers to the questions you're asking?
... codec that can be used across devices / operators? ...
... i lieu of that, a standard for describing the stuff? ...
... all you asked for is possible, but it gets harder with every format you add ...
... don't see us getting to that thing unless we get better interoperability ...
... pushing to device manufacturers if we want to see this move quickly ....

Ansuman: need some guidelines ...
... when to use what ...
... various standards ...
... scenarios, guidelines ...

Strategic thinking for Web Video, Franck Denoual (Canon)

[Slides, Paper]

Franck Denoual: .... defining "acceptable quality" out of scope
... might have different kinds of interactions for different kinds of content ...
... need ui abstractions ...
... video any time, any where: say, you get a message from a friend on your mobile and want to view the video that's linked ...
... video life-cycle ...
... rendering with seamless integration in to web pages ...
... concerning quality, content adaptation ...
... producer needs more control over quality --
...example: protocol with fragment identifiers for parts of a video ...
... so much about content adaptation, now interactivity ....
... same level of interactivity as with a simple image
... expect video to resize with web browser window ...
... "interesting regions"? ...
... also temporal interactivity: ability to skip at any specific time ...
... temporal links? ...
... to conclude, there's success, but also room for improvement ...
... interaction & quality ...
... enriching control of video objects inside HTML ...
... add metadata, could maybe bring added value to end user ...
... ie a solution would be open, scalable video format ...
... quality improvement! ...
... negotiation? ...
... characteristics of video? client requesting video according to display characteristics? ...
... could imagine matching ...
... if no match found, maybe transcode data ...
... and resort to content adaptation or proxies ...

Hitachi Position Paper: Video on the Web, Robert Freund (Hitachi)

[Paper]

Bob Freund: sorry, no slides
... remember an early Tim Berners-Lee presentation ...
... he clicked the underlined thing, and it all changed ...
... we're not there for Video ...
... we have video everywhere ...
... for a company like Hitachi, "who's interested in Video?" -- and everybody raises their hands ...
... deep-seated feeling that needs of technology growth has to follow social needs of people ...
... social needs of video ...
... lots of video in Japan, on portable devices ...
... baseball very popular ...
... there's instant replay of last inning ...
... they charge -- but tremendous desire ...
... boyfriends, girlfriends, birthday parties, long commutes are drivers ...
... to spend time ...
... lots of cameras out there -- all the phones have video cameras ...
...
... Back when Tim talked to OSF, one of the compelling things was that everyone can be an author
... put the ability to produce video content in the hands of everybody ...
... everyone who has access to a camera and a PC ought to be able to produce good content ...
... access, address every frame of that video using URIs ...

<Chris Lilley> +1 to addressability and to third-party access via linking

Bob Freund: URIs in and out ...
... how are people going to do Video mashups ...
... text mashups, app mashups, ws mashups, video mashups can't be far away ...
... how do we get the tools for that ...
... do we give authors the right to protect their creations ...
... most professionally designed web pages have a copyright assertion ...
... what are these people claiming ownership of? ...
... text, appearance, behavior? ...
... metadata causing behavior? ...

<raphael> +1 to address any part of multimedia content on the web using URIs, temporal frame and region in images ... recurrent theme in all presentations :-)

Bob Freund: battle for control over user agent rendering has been won by page owners ...
... more and more control over how user agent renders data ...
... need to do the same thing with video ...
... so that the mother, the father, the girlfriend, the boyfriend ...
... can make videos of that birthday party ...
... add metadata ...
... how do content owners express their rights today?
... bookmakers for a long time said "I assert ownership" ...
... DRM guys trying to take models from broadcast industry, recast on the Web ...
... that failed ...
... we're not the 7pm time slot in the NE corridor, we're the world ...
... it's ok to say "I own this" ...
... "I own it" good ...
... "you can only watch it 4 times" bad ...
... books stay for a long time till they wear out ...
... maybe digital things should wear out ...
... limit DRM for long extended use ...
... needs to be embedded ...
... metadata, links in and out, user agents and whoever wants to do search ...
... should be able to search, find your expression of rights, catalog it ...
... please solve this. ...

Ian Burnett: You talked about embedding everything into the video ...
... scalability issues of that? ...
... if everybody keeps packing information into the video, it might become huge ...

Bob Freund: it is huge to begin with ...
... if you have URIs to address things, you can address some bits that aren't shown ...
... most of the metadata will be text or XML ...
... that's verbose, but video is much more verbose ...
... watermarks could make it maybe 20% bigger ...

Glen Goldstein: There are some logistical projects with embedding metadata in video ...
... end-user comments, commentary tracks, tags ...
... not in a position to crack things open and insert? ..
... in that case, you're stuck with XML sides on the side

Bob Freund: didn't mean to say all metadata should be embedded ...
... rights assertions should be embedded ...
... should be able to associate with any URI ...
... associate metadata to it ...
... but bits that you want to preserve must be persistent, preservable, authorable ...

Stephan Wenger: Impression that anything since MPEG2 has some support for freely formed content ...
... that isn't really video bits ...
... to be embedded with Video ...
... what does that tell you?

Bob Freund: If I add something arbitrary, I'm not sure you'd understand

Stephan Wenger: There is a mechanism to put copyright statement in there, I proposed it.
... it's not deployed ...

Bob Freund: ??
... the web had a certain degree of completeness, then it took off ...
... video isn't there yet ...

Chris Lilley: you mentioned something about 3rd parties and mashups ...
... can't rely on markers in the video ...
... I might want to take just the candle out ...
... of the birthday party ...
... you might not have put markers there ..
... I like your vision, but maybe we need a more fine-grained addressing scheme ...

Bob Freund: think you need more than just fragment identifier ...
... being able to address down to time marker or frame ...
... is probably a natural thing for something that is temporal like video ...

Chris Lilley: I could also compose things ...
... rights stuff -- you might not like Mickey's ears ...

Bob Freund: If I write a book today and somebody takes that book and goes through with a marker ...
... to highlight what they like ...
... that's fine, that's tough on me if I don't like it ....
... if you want to put a mickey mouse head on me, that might be defamatory, not sure what the law says.

Steve Bratt:People should check out W3C's emerging "POWDER" spec -- allow authentication, description external to any resource - http://www.w3.org/2007/powder/

<Michael Dale> annodex temporal urls: http://annodex.net/TR/draft-pfeiffer-temporal-fragments-03.html

Bob Freund: the web works by virtue of the IRI
... to make Video a first class member of the Web, you have to have that ...

Sony Position Paper for the W3C Video on the Web Workshop, Bruce Fairman (Sony Electronics)

[Slides, Paper]

Bruce Fairman: Sony involvement in standardization ...
... how does home network (which is potentially independent of the web) ...

Bruce Fairman: that's different experience than the Web experience ...
... what's relationship to existing consumer electronics standards?
... how will home network users access that information ...
... ubiquitous is a point we should understand ...
... ubiquitous where? ...
... home network vs wide world? ...
... number of presenters will address that ...
... DLNA has architectural model, 1000 pages of guidelines ...
... media formats ...
... web vs commercial purposes ...
... video and a/v ...
... important that these formats be unified ...
... DLNA scoped for home network, not the world wide web ...
... different world ...
... client/server model ...
... UPNP to discover devices ...
... link protection to protect content on local link, DRM big issue...
... DRM is going to remain an issue for individual providing video on the web ...
... not in the same realm as commercial content providers ...
... DLNA has product certification and logo process ...
... goal is ubiquitous environment for home network, not WWW ...
... as an idea of what scale is ...
... 240 companies involved ... ongoing effort, several years ...
... home network piece of larger ubiquity ...
... larger ubiquity is what we're interested in ...
... home network, TV viewing experience, tied into the WWW ...
... viral video content ...
... not part of DLNA's considerations ...
... may become so in the future ...
... CE environment vs world wide web ...
... broad interest ...
... considering this environment for larger Web environment is important ...

Larry Socher: is there any reason to have different protocols and standards?
... want to do search, discovery, digital rights, across everything ...
... why different standards?

Bruce Fairman: content standards should fit ...
... but in terms of network ...
... we don't necessarily have a computer ...

Larry Socher: DLNA performance isn't that good
... on most implementations that are out there ...
... metadata indexing ..
... user-generated content doesn't have metadata ...
... I think we're converging on similar problems ...
... media as content from the WAN on the LAN ...
... could be direct, could be indirect ...
... sth that both parties need to consider ...

Philippe Le Hégaret: on the subject of metadata, I was in NY yesterday, panel on video search engines.
... was approached by video producers, what can we do?
... my answer was, add as much metadata as you can

[break]

Video on the Web, Kevin Lynch (Adobe Systems)

[Slides, Paper]

Kevin Lynch: Fan of video tags
... Demos scripting capability to more seamlessly embed
... Flash being adopted quickly

Kevin Lynch: Open to move video standards forward to support deep linking
... deploy on Web in a year; important for the Web
... encourage you not to think of Flash as alien to Web
... virtual machine in Flash is open source now
... hosted at Mozilla
... same engine in Flash and HTML and Firefox
... we are aligned to push Web forward
... Different video formats
... what we see with Adobe's video production tools
... they want H.264 as the format; see this as standard embedded across consumer electronics
... it's in set tops, hardware encoders
... Adobe has put H.264 into Flash; expect adoption in PCs next year
... Adobe's view is that H.264 is good bet
... HD is also coming to Web
... YouTube is seen as low quality video
... HD Gallery demo in Hulu
... see that quality is improved, can run full screen, live stream of HD quality
... this will be a revolutionary increase in quality
... Web evolving inside browser and desktop
... now making apps using Web technologies
... have desktop integration
... consider video not only in context of browser, but also in context of the device
... become a mainstream computing experience
... we're working on an effort called Air
... media players should be built with Web technologies
... references slide of key issues such as Searchability of indexing
... general problem of indexing rich content into computer
... not simple browse one page to another; changes state locally in ways that don't represent itself in URL
... URL goes to the whole app
... how do we do it in the whole experience; time code; technical challenges

Jeffrey Campbell: What's happening with Flash in mobile environment

Kevin Lynch: Flash lite
... partner with Nokia and Samsung, NTT to deliver services to those players
... Flash lite 3 supports video
... Nokia was first to ship devices; great progress

Eric Hyche: ISO container format?

Kevin Lynch: Yes, ISO MPEG-4 container format

Guillaume Olivrin: Can you support XML in Flash?

Kevin Lynch: haven't seen that too much

Dan Connolly: 90% of desktops?

Kevin Lynch: yes. Flash has more reach than any other technology in the world (more than Windows)

Doug Schepers: You are in favor of having video element in SVG, HTML?
... if you embed a Flash video; what is the format you would expect?
... what can users expect to experience; interactive; what are you endorsing?

Kevin Lynch: video tags should point to a variety of video formats
... whatever is delivering experience into video tag
... expose the DOMs across these environments; how to bridge HTML world and video
... "how to connect the embedded video to the DOM" as the big question

Standardized multimedia elements in HTML5, Kevin Calhoun, Eric Carlson, Adele Peterson, Antti Koivisto (Apple)

[Slides, Paper]

Kevin Calhoun: Now is time to take momentum and bring them to fruition
... there is interest, knowledge, and time to work out issues
... goal is standardization of markup
... not to supplant SMIL and SVG
... convenient syntax

<Dan Connolly> "codec agnostic design" didn't work for SMIL. I wonder why it would work now.

Kevin Calhoun: depends upon user agent that is being targeted
... be CODEC agnostic
... scripting API needs to be rich enough to accommodate
... that's the goal
... draft spec, work in progress we'd like to show
... webkit.org demo
... simple case of a video element
... two attributes
... one requests that its built-in controls be displayed
... a second attribute that requests the video is played as soon as request is made
... video is just another element
... scaling, resizing, reflow works
... some links from text into the timing of the video
... call a JavaScript function
... link into time stream in that way; this is a method
... other goal of HTML spec for video is that it be interoperable
... like to see it work elsewhere as well
... FireFox supports
... video element has two source elements
... video MP4 type, and OGG type
... we think that ability to select is important
... compositional behavior and rich scripting
... we went to Apple design people
... asked them what kind of page they would build
... they built demos
... demos that video can be composed in Z order
... CSS animations; includes positioning
... controls implemented in HTML and JavaScript
... see one subtle point

<Chris Lilley> (not smil animation of css, but animation in css syntax)

Kevin Calhoun: composition of video of things with rounded corners; the Holy Grail of composition
... with video as native element, we can do things like [demo]
... PNG graphic with alpha; video blends with the alpha smoothly

<Chris Lilley> compositing with alpha PNG

<Dan Connolly> this is pretty inspiring. PNG alpha composed with video...

Kevin Calhoun: mouse over items in bottom
... click on images brings up a featurette

<Chris Lilley> (could do the same in SVG naturally)

Kevin Calhoun: tilted at run time via CSS

<Beth> Are these demos available on the public internet?

Kevin Calhoun: CSS animation; dismiss featurette; fade is accomplished by a CSS transition
... if we make video in HTML, power of scripting allows us to accomplish some cool things
... move on to questions that may arise
... we're not authorised to talk about shipping products or IP rights
... what we can answer is whether standardization effort is complete
... next slide lists issues of concern to Apple
... we do want to achieve an open and implementable standard
... for codecs and formats
... want the strongest possible "must" requirement
... have to go through several stages of work
... go through container technologies
... candidates for these technologies need to be evaluated
... demonstrated their suitability and remove barriers to adoption

<Dan Connolly> his sketch of codec standardization makes a lot of sense; I hope I don't have to chair it, though!

Kevin Calhoun: our effort is to engage in this process
... while that process is going on, other open questions can be addressed in parallel
... other worthy attention to other issues such as accessiblity of timed media
... textual captions, audio commentary on the video?
... please engage in these questions with us
... other points are in our position paper
... open floor to questions
... and with other Apple colleagues

Glen Goldstein: presence of multiple source tags is key
... I don't see a bit rate or pixel dimension
... need to see dimensions and see the capabilities and connections

Kevin Calhoun: agree

<Dan Connolly> (one idea is <video src="foo.smil" />, but tighter integration prolly makes sense.)

Kevin Calhoun: whether content should be evaluated based on bandwidth
... Quicktime ref movies; select alternate content based on bandwidth; we'd like to find a solution

<Dan Connolly> (yeah... quicktime ref movies and SMIL... I blogged about that in w3.org/QA/ )

Jack Jansen: you seem to be reinventing things that SMIL already does
... why re-invent in HTML?

Kevin Calhoun: we want to make it possible for time media to be implemented in HTML documents
... no plan to supplant
... we value video tag in HTML to display SMIL based content; no intention to reinvent what already exists

Silvia Pfeiffer: multiple video tags
... different formats have different functionalities and capabilities

<Dan Connolly> (I think apple observes that it's easier to take the community knowledge in HTML and infuse stuff from SMIL into it than to ask authors to sorta start over with <smil> at the top rather than <html>. I dunno, though.)

Silvia Pfeiffer: what you suggest may make sense for mobile, this bandwidth, then client and server can negotiate
... but a more fundamental problem with this approach
... encapsulation formats
... start playing back time offset (I'll talk more later)
... link to video tag, not at beginning but to 20 second time set
... Flash you need a server extension to do that

<Dan Connolly> (HTML is certainly plaing a kitchen-sink role, which doesn't make the job of HTML WG co-chair any easier.)

Silvia Pfeiffer: all file formats require a server extension to do this
... isn't there a problem doing these multiple source tags; restrict functionality of what the user gets?
... may get a different presentation
... person authoring page will have a nightmare
... HTML page
... Real for Linux, Windows for computers, etc.; that was a nightmare for content producers

Kevin Calhoun: goal is to let content authors to make choices
... provide codings for the content
... limit number of features used
... opportunity to choose richer set of features as content requires

Chris Lilley: Great demos
... need for videos with overlay
... surprised to see it being done through CSS extensions rather than SVG.
... hope we can continue to see new things
... Opera was using SVG to do things

Adele Peterson: you will continue to see improvements
... keep eye on Web kit blog
... for the CSS transformations
... for people unfamiliar with SVG this is a good way to get started

Thoughts on Future Requirements for Video on the Web, Eric Hyche (RealNetworks)

[Slides, Paper]

Eric Hyche: I can't do it personally, or I can't do it easily
... I want to add value to content that I don't own
... want to modify content and redistribute it
... Philipp Pohlman
... go frame by frame to view scenes and relate to characters
... I wanted to add value to that content and share with someone else
... add comments to the track for example
... not just a playback
... maybe advance frame by frame and start back up at certain points
... or pause to add annotation
... all these things I want to add and redistribute legally
... another thing is to use richer meta data; more adapted to video
... for example, funny scene on TV show, "The Office"
... then send this scene to someone
... richer meta data ...show me clips when characters are on screen at the same time
... this requires new types of metadata
... what we store is static in the timeline
... I cannot search for where Nichole Kidmann is on same screen at same time with someone else
... Web Video; we're not talking about video on the Web by itself
... people want to transfer the video to their own devices
... my ten year old asks how to put video on to her player
... RealPlayer11 may have some interesting features
... browser extensions that show when video is playing on Web, and when to download it
... download video to desktop and play offline or transfer to your device
... secondly, that's for things that are downloaded via HTTP
... there is live video transfer
... or RTSP server
... I want to save those, too
... RealPlayer lets you save those to your desktop and replay
... How W3C can help - standardize in your domain
... standardize on temporal and spacial meta data
... a DRM agnostic rights description mechanism
... How W3C can hinders
... if HTML video tag is incompatible with SMIL; or not future proof
... and not support synchronization between multiple thigns
... or should not standardize codecs, file formats
... can be hindered by standardizing outside W3C's natural domain, i.e., codecs, file formats, delivery protocols
... show you demo
... of RealPlayer 11
... go to YouTube
... click on tag
... downloads to my desktop
... while it downloads; show something else
... video clip trailer
... as that is playing, I can record it
... I don't have to watch the whole thing and save it into my library
... and then look at it locally in my disk and play it back
... plays back live stream of what you saved
... now YouTube download is done
... go back to my library
... saves it as an FLV, not another format
... invites questions

Julie Lofton: with recording aspect of bringing to desktop, what are the rights issues?
... have you thought about it; what are your solutions?

Eric Hyche: yes, we did
... we think of it in two ways
... the rights to play back needs to be thought through
... not rocket science to pull video to desktop
... rights to what user can do has to be worked out
... if someone wants to restrict the playback, the technology exists
... content providers are putting out in format that's not protected
... we have a solution for content providers to not allow that to happen
... if you mouse over it says 'this video cannot be downloaded'
... that question of how they do that, is in the hands of the content provider

Julie Lofton: the solution of this video cannot be downloaded
... you'll only get to use it with YouTube or user-generated content

Eric Hyche: not nec; we have solution for content providers to decide
... that's their choice

Ian Blaine: why content people put content online is to then monetize through advertising

Eric Hyche:that's important
... protect rights holders rights
... preserve the right to play back, and also preserve the branding
... so it's not lost when people put it up there; we're working on things that address content owners rights
... to hold onto their branding

Rigo Wenning: Is Eric aware of ODRL submission to W3C which is rights expression language
... it lacks support
... there is some momentum
... or is this Creative Commons approach

Eric Hyche: yes, I remember rights submissions to W3C
... I didn't see those go beyond submissions at that time

Dan Connolly: You could look at POWDER working group

Position Paper for the W3C Video On The Web Workshop, Chris Double (Mozilla Corporation)

[Slides, Paper]

Chris Double: brower implementation
... some has been covered by Apple; we're all working together

Chris Double: find a solution to make things interoperate together
... demos I will show run on Opera FireFox Web browser

<Dan Connolly> "these are originally opera demos; I haven't touched them at all and they run on the mozilla browser"

Chris Double: what made HTML successful was that anyone could publish it; build tools
... not worry about patents, anything to stop them; as long as they had Web server and right tools
... remember Hot Dog
... we want people to do the same sorts of things with video
... how to build tools

Chris Double: tags, file formats to easily upload; with any Web browser
... not just Web TV

Chris Double: phone cameras, show to friends and family on blogs
... right now upload to YouTube and embed Flash
... have to deal with third parties, possibly give up rights to host and play video
... how to do this?
... dedicated mark-up
... video as first class object
... people should be able to copy/paste
... not everyone wants to do scripting
... do markup
... video needs to integrate into rendering pipeline of browser
... cannot be thrown into a separate window where browser has no control
... my slides are HTML; embedded instance of video here
... video source to file; controls, click on play
... demo that Opera put together
... How to build tools?
... legal ground you are on; want to use file formats that you have the rights to be able to distribute
... and not pay onerous royalties
... encode video from cell phone to Web format
... built encoder into system; need to make that simple
... issue we have to solve
... distributor of decoders have to solve problem of whether to make royalty payments or licensing for decodres
... hard for implementer to get it out there
... open source browsers; unclear what can be done
... people can build tools; DOM interfaces; build own player within the browser
... an embedded instance, uses player in JAVA Script
... anybody can write using simple interfaced defined by the DOM object
... WHATWG has a proposal; work in progress; it's on the public list
... multiple browsers have implemented the specs
... WebKit, Firefox, Opera; they interoperate
... author can expect it will render as they want
... should there be a baseline codec
... so a producer of video can code to that format
... this is important to have
... a baseline so you can produce content that will run across any platform
... if you need special features, then those can be plugged in
... but something in basic case to upload video
... avoid file formats "not supported" messages
... people don't like to read a user license just to see a video
... need to work out a way to deal with that
... demo of videos playing animated by JQuery JAVAScript
... very easy to do once we have the spec and standard tag in place
... head to future that allows this
... In Firefox we are using the OGG THEORA
... open source, easy to get, BSD license
... originally WHATWG had a should requirement to support OGG Theora and Vorbis
... need to examine requirements to see if they suit what we need to do
... we also need to analyze if this is the right choice; come to a consensus
... and come up with a codec that is the right option
... issue of patent risk for codecs
... submarine patents you don't know about
... companies like H.264 have already taken a risk; may not want to take on further risk
... need to work out
... DRM not that big an issue; people can do DRM; browsers don't have to implement just one codec
... source element to select particular codec
... we need a baseline codec
... summary, browser vendors are working together, implementing a standard that people have discussed
... WHATWG is starting point

Philippe Le Hégaret: References comment from Dave Singer http://lists.w3.org/Archives/Public/public-html/2007Dec/0044.html

Dan Connolly: you have integrated an OGG implementation; have you looked at implementing H.264?

Chris Double: Not at this stage; other people have to make their decisions

Dan Connolly: some people think there are patent risks around OGG format

<Chris Lilley> a software engineer viewpoint is also interesting in fact (e.g. ease of implementation)

Dan Connolly: offers we could use W3C patent process to help fix this problem
... I hope to write something about this
... please talk to me if this is of interest

Keith Lantz: What hope does Theora/Vorbis have, given the tidal wave of industry support for H.264?

Chris Double: Good question, I don't know the answer
... the idea of contributing to the Web is key
... to having things take off
... without making those licensing payments
... it may be a niche format for user-generated content

<rigo> I think W3C has to make sure that we do not get back into the gif - situation where some rights holder was periodically going after some users creating waves of uncertainty

Web Architecture and Codec Considerations for Audio-Visual Services, Stephan Wenger (Nokia)

[Paper]

Stephan Wenger: my paper has generated media interest
... please view in context of standardization only
... and do not stick it to product statement
... This talk is about codec choices specifically
... perhaps similar problems come up with transport protocols
... Basic message is that good idea if W3C

Stephan Wenger: would stick to its center of expertise and not venture into areas where other committees view as their expertise
... I'm not commenting on terms such as "free", "proprietary" or "open"
... these terms are interpreted radically different between MPEG and Open Source communities
... my choice of terminology regarding proprietary may not have been most sensible for this audience
... but it is from groups that I originally come from
... HTML draft 5 spec; generally it's fine and dandy; we think there is a need for a video tag
... but up until yesterday there was an offending sentence about video tag
... reference to codec OGG Theora
... no other codec was mentioned
... why that was included in original contribution, the W3C has a clearly expressed
... and consistently followed the idea of standards that do not bear royalties in any form
... generally that is a good thing
... everything that's an HTML doesn't have to be royalty free
... and still be in compliance
... in practice, whatever W3C has been doing in the past has been free, or perceived as free.
... against advice from companies that have a lot of experience, and have spent time, energy and have commercial interest;
... WHATWG people decided to put in that sentence into HTML5 draft spec
... we cannot shed light on the details of our advice against the use of OGG/theora.
... A W3C Patent Advisory Group (PAG) is triggered when...it's mandatory, therefore we cannot
... so we cannot say anything to a PAG
... The spec, as it was written a few days ago, may force us to implement something we don't want
... and we cannot use procedures of this organization (W3C) to argue with it on patent grounds
... that's why I wrote this position paper, and why it drew so much attention
... Theora is inferior in all aspects to modern codec
... from its technical starting point, it's close to H.261
... so technically the performance is like H.261; we can do better; need to be future-proof
... I wish I could throw 7-digit numbers at you
... I believe the copyright and source code is a minor issue is minor compared to the patent issue

Stephan Wenger: a number of initiatives out there (to implement H.264 in the open)
... where many can do open source and include code
... JVT (Joint Video Team) committee (ITU, MPEG)
... licensing terms for the reference software basically says, 'don't sue us' — no other restrictions.
... other reason is that we as a product company
... with scary restrictions on what we can build in our products
... most likely would not resort on what the open source community resorts to (e.g. an open source implementation of any codec)
... we would do the implementation from scratch as we do with other things
... let me get broader here
... I perceive a clash of ecosystems

<Dan Connolly> (I think performance was the most relevant restriction he mentioned)

Stephan Wenger: perhaps not limited to this
... Web standards are free
... digital compressed video area standards are NOT free
... royalties are there for content and distribution and playback of the content
... there is no "free" today in digital video
... if you want to evangelize the world towards “free” digital video, the
... Web ecosystem would have to convince the other ecosystem
... to follow your approach
... people are making money on content
... can you do that? how do you do that?

<Chris Lilley> I thought that the Theora codec was developed by On2, originally as closed competition to MPEG-4's H.263 (DivX) and H.264 (AVC) codec?

Stephan Wenger: I know it would create havoc for both ecosystems
... I don't want to go through the pain
... Web should stick to Web stuff
... leave aspects other than that; such as codecs to others
... can the two ecosystems co-exist in products and standardization space?
... we can see that the only sensible choice today
... What would be the only true candidate? pull out someone in the street
... and they would say FLash is choice for video on the Web
... Flash is proprietary and coexists well on the Web
... maybe too well for some of us
... In specification space, you do your stuff, and the video compression engineering groups do theirs
... you shouldn't go recommend stuff you're not so familiar with compared with those pepole
... there are some alternatives
... H.264 for HTML would make me happy
... but it would make others not happy
... one alternative is what is in HTML5 draft right now: list of requirements
... at present no codec fulfills all of them
... as there is a list created by people from
... people from either camp; possibly the licensing side

Stephan Wenger: licensing is based on commercial requirements of the right holders
... so changing their commercial environment, that's one possibility
... W3C does some evangelizing but not without much review
... second aspect in the paper
... codecs which are going off patent
... H.261 is widely believed because of their age, 1988, that they would be out of patent coverage, so use those
... the third option is to look at commercial and technical viewpoints and come up with a conclusion

Dan Connolly: thanks for being the messenger who gets shot; thanks for being here and being accountable

Video on the Web at the Walt Disney Internet Group, Elisabeth Freeman, Steve McQuade (Walt Disney Internet Group)

[Slides, Paper]

Elisabeth Freeman: ESPN, ABC as well as Disney brands ... video everywhere
... each brand has customized video, ads, controls, etc.
... 1st TV company to offer full shows in Web
... need customized play (for experience, tracking, reporting)
... mostly for free with embedded ads ... new interactive ads
... users remember ads better (development of ads more expensive, costing to advertisers is different)
... subscription content in iTunes
... This year started user-generated content
... "virtual rushing" of fraternity or sorority
... they have to moderate for both content and copyright
... participated in user-generated content principles process
... http://www.ugcprinciples.com/
... mobile .. haven't fully utilized it
... Challenges
... cost of distribution
... performance and QoS
... piracy
... how will video on web play into piracy problems
... advertising (no stds for impressions, tracking, reporting, working on this
... mobile (trying to stay on top of new tech., and standards)
... that's it.

Larry Socher: Disney Mobile .. voice centric .. how will it come together with video

Elisabeth Freeman: Dis Mob folded
... still important for future

Bob Freund: no standards for counting impressions for video advertising

Elisabeth Freeman: Working with IAB?

Julie Lofton: what kind of rates for CPM, interstials

Elisabeth Freeman: Can't say ... but they are still doing video on web

Glenn Goldstein: products look good. Installing new players may be a deterent. <video> should make easier

Elisabeth Freeman: Flash is de facto standard. But they will explore HTML5 video tag

John Toebes: Content put on shortly after on TV. Why:

Elisabeth Freeman: Others want to see if quickly because they couldn't see it when on broadcast (this is primarily an issue in international markets)

Potential for Internet Video Monetization, Shashi Seth (YouTube)

[Slides, Paper]

Shashi Seth: Average CPM in industry is $10-15 on video
... 120 on YouTube have been signed by record label or broadcaster

Eric Hyche: How does revenues from Web video compare to TV?

Shashi Seth: 5%

Ian Blaine: Breakdown of CPM for premium vs. UGC

Shashi Seth: #streams of the 2 are about equal
... At YouTube, CPM is not lower or higher than UGC

Philippe Le Hégaret: "CPM" ?

Shashi Seth: Cost per Mille, 1000 views of an ad
... $12-18 CPM
... clicks are being counted, but not for payment

Medical Uses of Video, Neil Izenberg, M.D. (Nemours Center for Children's Health Media / KidsHealth.org)

[Slides, Paper]

<Dan Connolly> "color fidelity is essential". wow. that was a great slide

Neil Izenberg: Interesting data about value of Web, uses of Web, med image and data file sizes, what is important regarding Web technology
... Meta tags
... Video considerations from his slides:
... Issues of file size, image resolution, color fidelity, security, and authenticity are fully addressed.
... Video can be annotated (metataged) in some way, so users can share and search. Medical ML?
... Perhaps resolution can be adjusted during use
... Perhaps video can have information built in that allows construction of creation of 3-D image, possibly using edge detection, luminescence profiles, or other data.

Bob Freund: Colors

Neil Izenberg: color is a critically important problem in diagnosis
... would like standardized colors chart accompanying image

Don Bruztman: How important are stable open stds since videos will become part of lifetime medical records?

Neil Izenberg: Likes open standards

<Dan Connolly> "epoch is privately owned" did I get that right? who is epoch?

Neil Izenberg: proprietary aspects of existing medical records a problem
... watermarking and fingerprinting important to ensure no bad manipulation of images

Amit Sheth: What are opportunities for metatagging?

Neil Izenberg: explanations should be attached to images (links to lab data, explanations of observations, etc.)

Bob Freund: Anonymization?

Neil Izenberg: HIPPA (US regs regarding med privacy)
... teleradiology. now common for readings to be done overseas
... lots of potential problems

Video as a first-class object for promoting scientific research, innovation and training, Sebastjan Mislej, Davor Orlic (Department of Knowledge Technologies and Center for Knowledge Transfer, Jozef Stefan Institute)

[Slides, Paper]

Sebastjan Mislej: Lectures by famous people available all over the world
... 800 researchers from IT, phyics, math, etc.
... 14 people actively working on videolectures.net
... interfacing with other IT research and departments
... online video knowledge repository
... free, open, no commercials
... aim to cover all knowledge
... Content providers
... EU reserach projects, cooperation with others in many countries
... Why is this project unique
... procedures to select content (editorial board), use of machine learning and data mining, backbone in many major research projects, publicly available

Dave Orlic: Functionality and structure
... YouTube like things, tools for lectures (CM, course buildling, ref material), direct access to lecturer comments, self-course building
... Formats: H.264, FLV, WMV, MP4, Ogg
... Stds: Dublin Core, Learning Objects Metadata
... Data structure: CERIF (exchange research papers, video, etc.)
... DJANGO- python
... http://en.wikipedia.org/wiki/Cyc

Jack Jansen: Interest in annotation?

Dave Orlic: Yes, and some semantic structure around this is important

Jack Jansen: anything now?

Dave Orlic: Yes, preliminary now

Philippe Le Hégaret: Web site providing an interface for transcribing, etc. ..Wiki-like
... one item was translated into 30+ languages, and was linked from 2000 weblogs (85% of them non-English)
... http://www.dotsub.com/films/inplainenglish/index.php

Is Video on the Web for Sign Languages, Guillaume Jean-Louis Olivrin (Meraka Institute)

[Slides, Paper]

Doug Schepers: is use sign language on movies, would not want to embed video in video
... so how to put this content out?

Guillaume Olivrin: Up to user to put overlay where user wants -- need control (like u need control on things like font size)

Julie Lofton: I speak fluent American Sign Language. American Sign Language is not understood in other countries. What about other sign languages? What about MPEG7?

Guillaume Olivrin: True. More than one SL. Don't want to std. on one.
... Video channel should be able to carry many SLs
... cannot just narrow to one set of rules... need to have broad ability of video to play back all
... Did create an MPEG7 profile for SL
... created annotation that is both spatial (speaking about this person "here") and temporal

Video Requirements for Web-based Virtual Environments using Extensible 3D (X3D) Graphics, Don Brutzman (Web3D Consortium, Naval Postgraduate School, Monterey California USA)

[Slides, Paper]

[no question]

Discussion

Ian Blaine: Any topics people want to discuss?
... Metadata one of the biggest
... codecs, RF of interest. Consumers don't care as much
... What are challenges around metadata
... let content describe itself, and let other describe it
... search, personalization

Rigo Wenning: company came after users in arbitrary way over gif. can't let this happy in video
... Policy Languages Interest Group is discussing privacy, access control
... description metadata an issue

Jason Gaedtke: Paying for metadata, but still cannot find authoritative source
... data duplication, poor quality
... lack of standard for naming assets
... as many as 50 copies of a assets may be released
... did analysis of content to look for common assets
... do others see this? is there an easier solution

Mark Kortekaas: CBS: Industry has does not have consistent tagging scheme even within a company
... repeats of show treated as new content
... 2-week lifespan of content

Jason Gaedtke: Is this a cultural or workflow problem?
... would a std be helpful?

Mark Kortekaas: chicken and egg
... film is different than sports, etc.
... need to grapple with why we want to retain data

Michael Wise: Lots of difference across their properties too
... look for metadata stds that work on Web
... different from production
... mapping from internal production, and the Web

John Toebes: Cisco: Production metadata from source (pretty authoritative), plus large body of UG metadata
... look at both to add value to video on Web

Paul Bosco: Are other stds bodies working on metadata?

Keith Lantz: The Augmented Multi-party Interaction Consortium. This isn't a standards group per se, but a consortium of mostly European university and industry research groups focused on facilitating real-time meetings. They have developed a large corpus of rich media data from meetings held in their instrumented meeting rooms, metadata to describe all that data, and a variety of tools and meeting browsers to find and retrieve the data.

<Chris Lilley> W3C have been involved in AMI (and AMIDA)

Keith Lantz: can help put W3C in contact

Amit Sheth: production metadata and user metadata, but also now also automatically generated metadata
... speech to text, closed caption conversion, video recognition
... generates RDF metadata automatically
... could be user-corrected, validated

Jack Jansen: Do we want to treat prod, UG and auto metadata differently?
... authoritative metadata may change less often than user generated

<Michael Hausenblas> we've put together an overview on multimedia metadata -> http://www.w3.org/2005/Incubator/mmsem/XGR-vocabularies-20070724/ MMSEM XGR

Guillaume Olivrin: need high-level metadata, then lots of details. May be value to many levels
... get limited, closed vocabulary for high levels

Identifying Spatial and Temporal Media Fragments on the Web, Lynda Hardman (Centrum voor Wiskunde en Informatica (CWI)/Joanneum Research, Austria)

[Slides, Paper]

Lynda Hardman: apologies for potential lapses into incoherence due to the latness of the hour in Brussels

Amit Sheth: Would you be able to add additional metadata with the same mechanism?

Chris Lilley: I would have used the named views in the svg example (see http://www.w3.org/TR/SVG11/linking.html#SVGFragmentIdentifiers). Is there a direct URI mechanism?

Lynda Hardman: you need to make the coordinate scheme transpartent to the person who want to annotate the image

<Chris Lilley> to rewrite the example from lyndas slide, the URI would be foo.svg#svgView(14.64,15.73,146.98,147.48)

Jack Jansen: Using a uri scheme to identify things in your media, will that break when things move?

<Raphael Troncy> yes, localizing means having a reference and a delta

Lynda Harman: Moving region over time complicates the problem.
... it ismore difficult, first worry about the temporal and the spacial and worry about the next later

<John> are there any such examples of deep linking into an image? i.e. a link to a upper left corner of the image.

<raphael> various use cases: localize logo on TV ...

Hyperlinking to time offsets: The temporal URI specification, Silvia Pfeiffer (Annodex Association)

[Slides, Paper]

Jason Gaedtke: Does the skeleton provide the mapping to get to the right location?

Sylvia Pfeiffer: That would change with codec...
... It is media format specific

Chris Lilley: you chose to implement something that has nothing specifically to do with the ogg format

Sylvia Pfeiffer: The adressing scheme for mpeg21 is different and previously done.
... it is an open specification and everyone can use it.
... it has been deliberately kept simple
... There is a recent discussion on the uri mail list on uri templates
... That way the server can provide a template that the client can adapt to

Kevin Lynch: It seems that we are struggling to deal with the nature of urls and not the way that local state in the client may be managed

Sylvia Pfeiffer: One of the features in the current proposals is that the URI changes as it is manipulated which provides a reference that can be shared
... We just wanted hyperlinks to work, but you are right

<Thomas Roessler> I'd note that, in order to keep client side state around, you can use fragment identifiers that are concocted using JavaScript

Sylvia Pfeiffer: The state of web pages is a larger problem which we are not eager to tackle

<Thomas Roessler> But +1 to this being part of a larger problem.

Eric Hyche: There are other timing things related to video playback such as speed or looping
... Have you given thought to specify these?

Silvia Pfeiffer: There would be some communication necessary. You can use http params
... or it can be another query param on your url

John Toebes: We are talking about deep linking into video where we currently deal with images without such support

Silvia Pfeiffer: Images do not have a time element. We would like to use the same sort of things for video
... that we ue for images
... At one point we looked at sing image maps
... it looks interesting for regional addressing

<Chris Lilley> portions (clipped regions) of images can be linked to

Video on the Web: Experiences from SMIL and from the Ambulant Annotator, A.J. Jansen (Centrum voor Wiskunde en Informatica (CWI))

[Slides, Paper]

Silvia Pfeiffer: You described how you can structure a video file.
... can you address into smil with a url?

Jack Jansen: sure fubar#frag

First Day wrap-up

Philippe Le Hégaret: what should be the next steps for the w3c?
... Is there anyone interested in describing their conclusion of the day?
... some of you didn't get a chance on stage, care to speak up now?

Rob Glidden: A lot of good things are being said
... A common theme is the value of RF activity in the W3C is high value for Sun
... we would like to highlight the open media commons activity

<Steve Bratt> http://www.openmediacommons.org/

Rob Glidden: which is a Sun initiative. We were told that we couldn't do it because there are existing organizations in this area

Rob Glidden: Vague claims of rights by others are maybe only fud or lack of knowledge

Rigo Wenning: The issue arounf RF is that it will need a very complex social process
... Truth does not exist, but we can use a process to bring the stakeholders to the table
... so that the problem can be properly framed
... We have already the experience of compression algorithms
... The patent policy process in the W3C gives u opportunity but no guarantee

Steve Bratt: I want to thank all of the presenters
... I picked up a few things for the w3c to do, bu it is early and will discuss more tomorrow
... There are a lot of technologies in process at the w3c (smil, xml...)
... We need to look at what may be missing. Many of these thing may need to be examined.

<Chris Lilley> I think we have rather shown that smil and svg have some things that the proposed html video element is missing

Steve Bratt: another thing is the identification of things in time and space.

<Chris Lilley> codec-specific addressing precludes conneg on video

Steve Bratt: There may need to be codec specific mechanisms to point into points of video time and space
... The question remains as to how we can specify that

Shashi Seth: Search engines used to use metadata to provide hints for indexing
... that creates the problem of spam since the metadata is separate and may be targetted at the search engines

Shashi Seth: There may be better signals that can be used instead of user generated metadata for indexing

?: The problem of spam to me is the model you use and how you layer the data or metadata that you use

[Second day]

b2b+social media=p2p, Matthew Patulski (Cap Gemini)

[Slides, Paper]

Matthew Patulski: Web-based media
... Business to business plus social media equals a person to person dialogue
... Basic premise of our work for last year, content, workflows, etc. [See slide]
... Capgemini provides services; person to person interaction is our strength as a company
... We deal with complex challenges, not unlike our discussion today
... we strive to put a human face on content
... commitment to social media is not just a marketing term of art
... we use documentary techniques for story telling
... two camera shoots to give it "you are there" feeling
... we like to think of our company as content rich but low key
... let subject matter expert's message come through
... not about loud music and anthems in our videos
... capture subject's enthusiasm for content
... [demo video]
... clips from six pieces produced over six months
... all about story telling that people want to watch and share
... we have developed our own Flash-based player
... Capgemini.com home page
... study about digital natives; tweens and how they use the Internet
... how it affects content and product trends
... embedded on our home page
... content is permalinked
... unlike CNN when video is not there a week later

Matthew Patulski: offsite, we have a branded player, like YouTube is embedable
... carries hyperlink back to CG site
... and title of offering
... if you want to share content, there is a socialization option
... three panels of choices to socialize via DIG, Technorati
... send to people, or cut/paste or embed code or put in your Web site
... features on built into onsite and offsite player
... soon to launch multi-player
... to develop content, it's about workflow and standards; moving multiple types of information to produce a video
... video in multiple formats (PAL), photos, etc.
... publish back out and put on the Web in other forms; PPT, DVDs, etc.
... whole spectrum of delivery points has challenges
... standard-centric workflow is a term of art
... externally, when I put a video on line, what's the metadata; what's the file format
... we have standardized
... global workflow; from Cupertino to Mumbai working on it
... centralized shared storage, shared assets
... standard-centric workflow enables repeatable results
... 2008 goals
... present larger format videos
... launch a dedicated screen player
... FLV; Screencast by Slideshare.net
... PPT presentations with audio annotation
... same kind of publishing tools, but content is different
... expand production network
... external push of our content
... we're on major sites now
... build relationships with clients and those who talk about work we do
... bloggers who talk about trends in supply chain, SOA,
... metrics and ROI are important
... building best practices for organizing communities of interest
... if we do a 60 minutes interview, we may only end up with 3 minute video
... what to do with the other 57 minutes
... challenges to organize it; with metadata?
... ID the bits; the annotation of it; taxonomy, folksonomy
... metadata is interesting for us here
... how that plays across broadcast and Web related content
... video bridges both those spaces
... we're running into demands of both, too
... big learning experience with challenges
... download presentation online and see the player
... see what we're doing
... we have a Capgemini YouTube channel, also BrightCove and DailyMotion
... Questions

Amit Sheth: For metadata are you using standard or controlled vocabulary?

Matthew Patluski: No we are not
... trying to use this effort to force that issue

Matthew Patluski: traditionally we have not had a taxonomy for our marketing material
... data related to managing video assets is astronomical
... forces the hand on this
... why I'm interested in this

Amit Sheth: If metadata were in RDF would it be more usable?

Matthew Patluski: yes
... database is key to a particular file
... one ideal scenario is why that transcript cannot be embedded into file of metadate

<Dan Connolly> (re metadata in RDF... see also GRDDL, where you can put your metadata in HTML or Atom or any other XML format and semantic web tools can read it like it was RDF/XML)

Matthew Patluski: we're not quite sure where to go; we're giving it thought

W3C Video on the Web Workshop, Jason Gaedtke (CableLabs)

[Slides, Paper]

Jason Gaedtke: we're a non-profit R&D consortium made of US, Canadian and some European cable members
... packet cable and VOIP ...new to this role
... was formerly at Comcast
... working on IPvideo
... consumers of technology you have developed at W3C such as HTML, SMIL, etc.
... I want to talk about how we get the bits to the consumers; perspective of network resource management
... peer to peer
... Abstract is we hear about network neutrality
... a difficult topic; strong opinions
... I'm sympathetic to a large, unregulated pipe; but also some economic realities
... first point is that not all traffic is homogeneous on the Internet
... next point is peer to peer; talk about TCP
... intersting data from bandwidth analysis
... some troubling conclusions about growth
... so other apps are not drowned out
... I don't have answers, but some areas to explore
... taxonomy of Internet apps
... three broad categories
... Real-Time Apps, latency sensitive
... bandwidth requirements can be a ping in a game to very high real-time apps like audio or video chat
... time sensitive to transmit
... down a layer is Interactive services

<Dan Connolly> (re real-time services where latency matters... check out http://ejamming.com/ )

Jason Gaedtke: you may wait two or three seconds; retransmission may make sense
... third category is where peer to peer agents live
... automated services; no human user; highly efficient, but also very greedy
... we'll look at network implications
... Web video
... Web servers, scale as you would any Web farm
... content distribution networks emerged from that model
... additional caching introduced
... Comcast built out a CDN of their own
... cache on network
... we'll look at IP address, if avail capability, and route accordingly; it's a hierarchy
... peer to peer networks is a decentralization
... highly distributed, self-organizing; notion of supernode
... decentralized vs. hierarchical
... I also spoke at W3C Tech Plenary about peer to peer; some merits here
... reduced barrier to entry; a great technology enabler
... counter point to this is that there is no such things as a free lunch

<Steve Bratt> Jason's slides from W3C Technical Plenary: http://www.w3.org/2007/11/Protocol_Revolution.ppt

Jason Gaedtke: someone has to subsidize in terms of bandwith, storage, and power
... CND/P2P Hybrid

Jason Gaedtke: the head of the long tail gets pushed into the network
... P2P protocols used in horizontal mesh architecture
... longtail gets sourced from the CDN
... efficiencies and reduced distribution costs
... can see costs drop to less than 10% of bandwidth, storage
... Let's look at a resource consumption example
... pulled directly from an IETF Internet draft

Jason Gaedtke: Bob Bricoe, BT and Univ. College, London
... simplified model, but has some salient features that make point

<Dan Connolly> Problem Statement: We Don't Have To Do Fairness Ourselves

Jason Gaedtke: 10Mbps; 100 subscribers on network, capped
... this is how networks are scaled

<Dan Connolly> this p2p discussion reminds me of something I read...

Jason Gaedtke: of those 100 subscribers, 80 buy high speed data to browse the Web

<Dan Connolly> Is BitTorrent Evil?

Jason Gaedtke: typical interactive users
... other 20 are P2P file sharers
... let things run in background
... analysis is during 16 hours of user activity
... P2P clients have free reign of network during the other hours
... Interactive users are active 10 percent of time
... 10% concurrency is overestimate
... P2P are automated agents, running all the time
... important point is the number of TCP connections
... Web email side, spawn two connections
... some hacks may spawn four or six
... contrast to P2P, will spawn 50 to 100
... Web sites, 10Megabits, get best experience; bump to 400 connections

<Chris Lilley> I suspect power users with multiple email accounts, and multiple tabbed browser windows, will have a bunvch more than 2 connections

Jason Gaedtke: obvious takeaway is average rate that each connection or user sees when there is not enough capacity
... and the aggregate volume over that period
... ten seconds
... P2P, 500Kilobits; 500:1 volume skew
... advantage of P2P is spawning more connections and more active all the time
... now let's look at four times the capacity
... actual experience of network improves

<Chris Lilley> http://www.cs.ucl.ac.uk/staff/bbriscoe/projects/2020comms/accountability/draft-briscoe-tsvwg-relax-fairness-00.html

Jason Gaedtke: same is true of P2P; gone from 500kbps
... only getting 30kbps improvement
... more capacity into network, more expansion
... skew is 1250:1
... somebody has to pay for this
... some of users are cost sensitive
... those people will go elsewhere
... if 30 Interactive and ten P2P go away, that doesn't nec. help
... P2P guys take back the extra capacity
... from service provider perspective we have fewer subscribers, less cost/benefit, greater network capacity, and more cost
... a rational operator won't add more capacity; ten P2P guys will consume it all
... if everyone would pay for P2P it would be fine; but others are subsidizing
... TCP congestion flow treats all flows equally
... address at that layer
... agents are very greedy; aggressive algorithms degrade the cost/benefit proposition
... if everyone adopts P2P, that could work; but otherwise other services would lag
... merky waters; economic, network, and public policy consideration
... upstream/downstream rate limiting
... promotes fairness
... Europe and Canada uses a tiering strategy
... application specific throttling is contentious
... this is a very challenging problem; balancing P2P vs. Interactive
... priority queuing; reservation based are other options
... finally virtual economies
... I'm interested to build the technical framework to build economic models into
... some organizations doing work here are IETF; P4P
... I wanted to present material and share thoughts with me
... difficult but interesting problem

Mike Wise: Enjoyed your talk
... We have chosen systems that aren't greedy at Turner
... more friendly algorithms
... why we did not choose Bit Torrent
... bad actor of household Internet effort

Mike Wise: compliment on your talk; there are some more friendly P2P

Jason Gaedtke: thank you; some legitimate service providers have embraced Bit Torrent
... have to be careful what you say
... there are other algorithms
... intent to make P2P protocols more network friendly

Glenn Goldstein: Can we get better through put; on demand streaming
... advance packets ahead

Jason Gaedtke: great set-up for next speaker
... something of an arms race; would still be congestion

Dan Connolly: Technical Architecture Group is trying to figure out where this fits on our list of issues
... I'm also W3C IETF liaison
... thanks for sharing this new information

Jason Gaedtke: it's new in the last week

Dan Connolly: next IETF meeting in March

Chris Lilley: To provide contrary personal experience
... I have two Internet connections
... VOIP and IPTV
... both 20 Meg down, 1 Meg up
... one family, one for work
... we have Bit Torrent stuff running and I get my limits

<Dan Connolly> one place p2p shows up in TAG proceedings is the agenda of our recent f2f meeting http://www.w3.org/2001/tag/2007/11/05-agenda

Chris Lilley: also have VOIP which I use a lot
... I don't notice much of an impact
... IPTV can run on many screens
... doesn't seem to have much of an impact

Jason Gaedtke: fair enough
... do you adjust the number of connections on the client?

Chris Lilley: I think 50

Jason Gaedtke: experiences vary; 20M is a nice connection, by US standards
... contrained more by upstream rather than downstream bandwidth
... perhaps your upstream is more constrained
... do you notice any quality of service issues?

Stephan Wenger: Vancouver discussion
... nature of critical comments; nature of things proposed required some form of policy control into the network
... goes against the IETF end-to-end ideology. The IETF (at least in principle) prefers limited intelligence in networks, whereas Bob's approach, well-working as it may or may not be, requires considerable intelligence in the core.
... I don't have a solution there
... it's rather unrealistic to assume that this concept will receive IETF blessing any time soon

Jason Gaedtke: I agree with you; apologize if I sounded overly optimistic
... encourage you to download the W3C Tech Plenary slides
... I admire Briscoe for proposing a solution
... instead of doing things instrusively, actually allow each flow to be accountable for itself
... begs what is use or policy

Dan Connolly: if the IETF community need a nudge; then I will do that

What would make the Web a true video platform, Loren Larsen (Move Networks)

[Slides, Paper]

Loren Larson: some great talks over the last day
... I'd like to start with introduction about what Move Networks is
... started seven years ago; we settled on how to do video
... idea was to make video on the Web like TV; it takes 30 seconds to watch
... some design principals were to use standards, use existing Web infrastructure as much as possible
... and deliver high quality at same time
... allow a TV-like experience on a variety of Internet connections
... loewr end PC
... works for on demand viewing
... deliver video for premium content publishers like ABC, Fox
... doing on demand episodic viewing
... Elisabeth (Disney) talk showed that yesterday
... move back and forth seamlessly
... show a demo for ABC.com
... starts at low bit rate, detects to see if it can go higher
... see pretty good quality there
... shows what we think is possible with HD quality streaming over the Net
... if we had wireless network, the stream would continue to adjust itself

Loren Larson: one of things in discussions over past day is video is getting better
... things we can put into the browser
... other things may not be a standard; things from third-parties
... at W3C, make sure others can plug into the ecosystem
... functionality
... player that plays some standard video

<Chris Lilley> I checked and underestimated torrent settings. max connections 400, max connected peers 100, max upload slots per torrent 25, may upload bandwidth 120 kB/sec no cap on download bandwidth

Loren Larson: so Web designer can plug in seamlessly based on media type
... some thoughts on metadata
... workflow we see, metadata needs to be provided up front off signal or file
... pushed into asset mgnt or content management system
... live channels; time-based play out
... how to publish that play list out in a standard way
... way to express those need to be standard
... way to describe the content; common interfaces will facilitate ubiquity
... if we can standardize metadata across the flow, would be seamless for end user
... metadata point
... when metadata comes in from publisher, it's authoritative
... cooperate within video ecosystem
... we're passionate to have content ratings standardized; what people are looking at in video
... ability to find play lists; how videos are grouped to find shows
... temporal links
... basic player controls
... allow page developers to provide environments to play video
... possibly something standard in browser
... policy language needs standardization; ability to express rights

Loren Larson: codec concerns

Loren Larson: concerns about putting a codec into a standard

Loren Larson: may hinder ubiquity
... could potentially limit the ubiquity of video and limit what we want to accomplish
... DRM
... questions?

Rigo Wenning: policy language is of interest to me
... we just opened a Policy Language Interest Group http://www.w3.org/Policy/pling/; please join in there

Rigo Wenning: in PLING we are collecting use cases, not just W3C specs; issue is much larger
... interested in what you think the policy language is and what it is supposed to be
... DRM workshop in January 2001 (http://www.w3.org/2000/12/drm-ws/workshop-report.html)
... which is still landmark; market was so fragmented
... only thing you can standardize is the expression language, not the enforcement mechanisms

Rigo Wenning: that still holds from this workshop
... nobody has given us the resources to do this yet

Loren: we would be interested in discussion offline

Chris Lilley: choice of codec and wrapper
... are orthogonal
... these are separate concerns ... so for example the matroska wrapper could be used with the theora codec

Loren: right; I wasn't intending to tie them together

Methods for Preserving Audiovisual Metadata, Michael Wise (Turner Broadcasting System)

[Slides, Paper]

Michael Wise: General problem problem of standards with metadata causing loss of metadata
... changing media formats causes a loss of data

<Zakim> ChrisL, you wanted to mention that codec and drm capability are orthogonal

Michael Wise: user metadata could be useful or could be a spammer
... difficult to advertise against content with poor metadata
... re-encoded linear content may or may not have relevant metadata. Usefulness is unpredictable
... downsampling content affects the amount of watermarking payload which can be carried reliably.
... 34 to 35 watermarking vendors that they have found in this space.

Michael Wise: watermarking could have a potential for use in basic usage filtering. Not exactly DRM.
... Will require industry cooperation in order for watermarking / fingerprinting to preserve metadata in order to avoid unpredictable results.

Jason Gaedtke: Issue of user contributed metadata having potential for spam, what about the counter examples of wikipedia and the process around it

Michael: If there is community added metadata around an asset it can be useful
... fair point

Karen Myers: User sending in a content from a news source, is the converse true where user contributed content can have a watermark,

Michael: If there is a watermark that can be applied in a trusted way, it could be used

Glenn Goldstein: The ISRC (International Standard Recording Code, http://www.ifpi.org/content/section_resources/isrc.html) code could be a good way to start.

?: Has there been any use of watermarking/fingerprinting in the search realm?

Michael: There isn't any open paradigm to help in that realm

Video Recommendation Based on User Preference on the Web, Soohong Daniel Park (Samsung Electronics)

[Slides, Paper]

Daniel Park: IPTV Forum includes both mobile and fixed/mobile convergence on their timeline
... OIF has a large reach to consumers so expects to be successful in efforts
... EPG works well for TV/IPTV, but needs a different tool to search the web content

Standardizing the Syndication of Media with Advertising on the Web, Mark Kortekaas (CBS Interactive)

[Slides, Paper]

Mark Kortekaas: Solution today is one-off custom flash players and it doesn't scale to 1000s of sites.
... not position of W3C to set standards for advertisements, but to enable the concept of an ad and let groups like the IAB set those standards.

Mark Kortekaas: Ad formats are going to change and come from many sources, but the player has to respect the ability to insert the ad in any form
... They don't break episode into clips, but simply put markers at the ad insertion points so that they don't have to re-slice the content.

Julie Lofton: Regarding rev share with content owners. They are so used to traditional license deals that they don't understand ad revenue shares. So you can't offer a licensing fee since you don't know what you can make. How do you base it?

Mark Kortekaas: Public information - deals with the current syndication product are revenue shares. They don't have third parties selling content.
... they have to have very good reporting in order to assign the commercials played and track it back to the CPM base.
... because they have a large database and large amounts of data they can generate the value.
... how the value of the content gets described is another topic. Others might have different business models
... and it is unusual for the TV industry.

Ian Blaine: How do we deal with mobile as it becomes more important as a platform? Do we think that OSs will mature to support SMIL?

Mark Kortekaas: Would like to hear answer from Nokia on that.
... But, thinks that the same issues will apply to mobile that applies to web

Video on the Semantic Sensor Web, Amit Sheth (Wright State University)

[Slides, Paper]

Amit Sheth: using existing standards with the semantic sensor web to create metadata
... shown with sensor web, but should be able to apply to anything else.
... can use a closed loop evaluating the metadata generated to adjust what is captured by the device.

Amit Sheth: Demo used content from police cruisers on YouTube as a source.
... used OCR to extract the metadata from the image.
... i.e. codes embedded in the captured police footage

Discussion

Elizabeth Freeman: How do we deal with creative commons people and the ability for content providers to control how their content gets used
... for example allowing people to use portions of scenes, not the actual characters.

Mike Wise: Needs to be a mechanism for studios to ensure that they will have a ROI on their investment in making a movie so that they continue to make them.
... looked at creative commons for their short form content and found some of the mashups to be very creative and generated interest.
... is it done in a way however that they can capture some revenue/build a sustaining business from it.
... mashups can help or hurt brands.

Stephan Wenger: Some content should not be touched in any way (e.g. advertising during a political campaign)
... but there are clear examples where content is made available for tinkering.

Stephan Wenger: there needs to be a solution that allows for controls over which type of protection is applied to the content starts with the content owner

Jack Jansen: This is a control problem, but becomes difficult to understand what is acceptable for any content.
... Subtitles might be ok. mustaches may not

Mark Kortekaas: Can see DRM might be useful in some circumstances. How do we recognize the valid uses of DRM.

Silvia Pfeiffer: html does not contain drm. If I want to protect it, I use other mechanisms. To control who can see it, don't use HTML, use other mechanisms. There are other formats which support DRM protection and development happening in that space
... that doesn't preclude us

Guillaume Olivrin: Once something is shared, it is shared with eveyone.
... if you publish some content with some form of advertising, do you want it to be with it for the rest of time.

Rigo Wenning: Most DRM take the opposite implementation from what publishers want to describe.
... we don't have a platform for people to come together and discuss issues behind control of the content.

Rigo Wenning: on one hand we have tightly controlled assets and on the other hand we have the free side which has different requirements.
... notes that IBM gives away patents for free to benefit from network effect.
... would content providers derive greater value from free-er distribution.

Ian Blaine: When people have compelling incentives, the system works. DRM doesn't have to be onerous, it just has to provide guidelines
... if people can't find the way to do what they want, they will bypass and it turns honest people into dishonest people.

Enabling A Richer Video Experience With Metadata, John Toebes (Cisco Systems)

[Slides, Paper]

John Toebes: Metadata used in advertising
... Types of metatdata: four main types
... Content attributes, content description, timed, hotspots

Amit Sheth: can see value of storing metadata outside video content, because it helps with mashups

John Toebes: descriptors of the content itself (length author etc.) are fine to be internal, but other metadata is best outside

Matthew Patulski: what about when 2 different metadata sets are referring to a common video file?
... how do we establish a common list of things to refer to?

John Toebes: things to keep inside content: rights, source, size/shape attributes
... descriptive aspects to be kept outside: human language (since translation possible),
... things that let other end users contribute metadata content, etc.

?: should that data be with the clip?

John Toebes: it should be associated with clip, which leads to need for URI identifier within clip that allows referencing

Bob Freund: when we search for text today, we find it (wherever it is)
... when we search for metadata, we find that but not the original source. so how do we find the original resource?

John Toebes: the answer is in the question, the metadata should be sufficient for deep linking to original clip,
... this is also a requirement for DRM, so that original context is not lost
... also enables quotation for reasonable use

Guillaume Olivrin: [question unclear]

John Toebes: confident that integration with HTML will eventually work well

Amit Sheth: client-server issues regarding metadata distribution... wants to plug RDF and microformats as relevant

John Toebes: context of metadata usage/presentation is important. just seeing metadata itself might not be very helpful, should also present relevant visual display (such as bookmarked video)

Handing Over the Keys: User-Controlled Metadata, Doug Schepers (W3C)

[Slides, Paper]

Doug Schepers: terminology: intrinsic and extrinsic metadata

Amit Sheth: can i use a different vocabulary/ontology for my metadata under your approach?

Doug Schepers: reasonable use case... what about conflicting metadata tho? resolution might be to return all alternatives...

Amit Sheth: might want to use multiple metadata vocabularies simultaneously

Doug Schepers: will add that use case

Guillaume Olivrin: intrinsic/extrinsic metadata approaches appear to be in conflict; what about archiving metadata?

Doug Schepers: perhaps some of intrinsic metadata could also be externalized

Guillaume: further issues when zip/archive compressed

Doug Schepers: authoritative pointer is important, regardless of downstream repackaging, might even be enabler for further monetization

Jeremy Condon: what about namespaces?

Doug Schepers: yes, some mechanisms might be needed for ranking each one's authority relative to others

Chris Lilley: yes namespaces are obviously important.. notes that applications don't necessarily need deep knowledge of each namespace but will have to preserve necessary information

Doug Schepers: OK, first version of this approach ought to be simple

Chris Lilley: as simple as possible but not simpler

Jack Jansen: metadata might be incorrect tho, would error handling fall out from namespace reconciliation?
... [gave example of error handling in GPS domain]

Doug Schepers: error handling might be outside of scope for proposed API

Sylvia Pfeiffer: this issue relates to deep linking of metadata
... there will likely be different ways of handling each. for simple example you gave, too-simple handling is unacceptable, correct?

Doug Schepers: outlined how there might be 2 ways of accessing information

Metadata For User Generated Video, Drew Major (Cisco)

[Slides, Paper]

Thomas Roessler: agreed that this is important issue, wondered about labeling, needs to be one of the use cases. I do not think this is an access control problem

Drew Major: video is a big case and is different than existing web in some respects

Ian Blaine: relationship between social networking and search in this context

Drew Major: simple use case from end-user perspective

<Thomas> also, the individual social network providers become trusted intermediaries. They might be the threat we're dealing with.

Larry Socher: yes but there is also the propagation issue for social networks, multiple people referring to same thing

John Toebes: don't fall into trap of solving search instead of metadata, the key is to get video metadata correct

Drew Major: sure, plus now is the time to think about it and prepare solutions before we are swamped by avalanche of content . what about a universal user database?

Sylvia Pfeiffer: closest thing today is FaceBook, prefers end-user control of such identifying information

Thomas Roessler: there are also aggregation issues, we need to be careful that our metadata aproach doesn't make the problem even bigger, policy languages interest group (pling) is working on this..

Nimish Radia: trusted federation of multiple parties is a solved problem already

Metavid & Free Online Video, Michael Dale (University Of California at Santa Cruz)

[Slides, Paper]

[presumed reference for Nimish is http://www.openmediacommons.org ]

Metavid wiki http://mvprime.cse.ucsc.edu/wiki

[demonstrating video annotation system]

Doug Schepers: it's awesome! rephrased as question, how awesome is it? :)
... how much is client, how much is server? performance seems zippy

Michael Wise: lots of client ecmascript with ajax

Architecture of a Video Web - Experience with Annodex, Silvia Pfeiffer (Annodex Association)

[Slides, Paper]

[showing demo on laptop]

Silvia Pfeiffer: required reading: Weaving the Web by Tim Berners-Lee

[surprising that only 8-10 people of ~40 said they read it]

Silvia Pfeiffer: offer to consider W3C or IETF standardization. my former employer, CSIRO, has refrained from producing patents around this, so it's patent unencumbered to the best of my knowledge.

Jack Jansen: what about SMIL modularization?

Sylvia Pfeiffer: modularization wasn't in SMIL when we implemented, were focused on specific problem when forming CMML, it is an open-source research project and they weren't prepared to switch in midstream

Jack Jansen: perhaps aligning in CMML 2?

Silvia Pfeiffer: actually we are in CMML 3 efforts now, (detailed some current issues)

Chris Lilley: also seems to align with W3C timed text standard?

Sylvia Pfeiffer: worked on that alignment last month, further review help is welcome
... and there are several research efforts working in this domain separately, some are adding annodex

Doug Schepers: audio?

Sylvia Pfeiffer: audio, video, any time-continuous data appears feasible

Eric Hyche: why not put all metadata in Ogg header?

Silvia Pfeiffer: metadata is relevant to portions of the stream and so it appears there, especially important in live context

Eric Hyche: OK, that explains CMML file construction since you are constructing a linear playlist

Silvia Pfeiffer: XML is hierarchical of course, but they also designed CMML to be sequential

Dan Shanori: what about business applications?

Silvia Pfeiffer: design intended to augment Web architecture and thus is business friendly

Future of Video and Next Steps Panel

Jason Gaedtke: some themes have been repeated... metadata...
... e.g. authoritative metadata for professionally produced assets...
... and on the other end of the spectrum, user contributed metadata... wiki style...
... and in between is the "prosumer" marketplace that looks like a sweet spot; e.g. it produces wikipedia, open source software

<Jack Jansen> When thinking about UG metadata quality people should read http://www.cs.dartmouth.edu/reports/abstracts/TR2007-606/

Kevin Lynch: I liked a lot of things at the workshop, e.g. Disney perspective; not from a technology producer so much. very valuable
... a few things:
... 1. deep linking into rich content: video, rich internet applications

Kevin Lynch: W3C is well-positioned to do that
... 2. DOM integration...
... [missed something] ... and SVG object model could be more integrated with HTML

Silvia Pfeiffer: the demo was using DOM in an integrated way. [?]

Kevin Lynch: 3. metadata, as Jason said. We have XMP, which is an RDF-based format. used for photos and such now; we're interested in taking that thru a standards process

Chris Lilley: I'm on a "create" mailing list and XMP comes up there; I'm interested to see XMP come to W3C

Kevin Lynch: good; let's talk about that
... 3. [or 4.?] the <video> tag...
... I think it's great, though I'm a little concerned with whether it'll be in all the browsers, ...
... though maybe the javascript shim is promising... or is it just more complexity? hmm.

Kevin Lynch: 5. the codec question. the bigger issue [than ogg vs. H.261] is Microsoft Windows VC-1

Silvia Pfeiffer: have you heard of Dirac? from the BBC...
... open source... based on wavelets. perhaps lower risk of submarine patents

Jack Jansen: [on BBC and patents.] BBC is trying to be sure Dirac remains open

Silvia Pfeiffer: I heard a January 2008 1.0 milestone re Dirac from BBC

Philippe Le Hégaret: isn't Dirac submitted to SMPTE?

Chris Lilley: yes, as VC-2

Jeffrey Campbell: I saw a presentation and Dirac's primary use case is archiving, so its compression ratio is not as good as H.264

Stephan Wenger: I hope everyone appreciates that BBC makes their patents under RF terms, though that doesn't say much about patents others might hold
... [technical summary of wavelet compression...]
... DIRAC uses wavelet compression for coding intra and the resildual. It also uses block-based motion compensation. History has told us that the smaller you make your blocks (for prediction and motion compensation), the better your codec gets. H261/MPEG-1/2: 16x16 blocks. H.264 and MPEG4: 8x8 blocks. H.264: 4x4 blocks. Wavelets are in theory and in practice the more efficient the more pixels one handles in combination. Therefore, combination of small block sizes and wavelets cannot lead to a very efficient codec.

<Jack Jansen> http://www.bbc.co.uk/rd/projects/dirac/overview.shtml

Stephan Wenger: H.264 has a smaller blocksize for intra-frame motion detection. Wavelets need bigger not smaller blocks
... that's at odds with using wavelet approaches

<Jack Jansen> They claim "2-fold improvement over MPEG-2"

<Chris Lilley> http://wiki.transmission.cc/index.php/FOSS_CODECS_OVERVIEW

<Chris Lilley> http://hdmasters2007.com/pdf/Presentations/HDM2007_Wilson-HDDC.pdf

Silvia Pfeiffer: I'm happy that by the time I presented, the other presenters and discussion had laid out the background so well

Matthew Patulski: my perspective here is consumer of these technical developments, I suppose...
... two huge problems: codecs and metadata
... when I look at Annodex work and [missed], I want it now; meanwhile, our infrastructure is mediawiki; I wonder how to integrate them
... I appreciate the rigor with which you all present your ideas and at the same time listening to each other

Ian Blaine: the discussion of video as not just something that you stick on the web... but that as time passes, the web around the video changes [ paraphrasing]
... one thing to highlight re metadata: inferred metadata...
... e.g. "turkey" and [missed] suggest "thanksgiving". that can have impact on search and all sorts of things.
... with the increased availability of video, discovery is becoming a nightmare

Paul Bosco: anybody who should have been here but wasn't?

Ian Blaine: Microsoft; difficult to have this conversation without them.
... and advertisers. I'd like to hear more from them, such as IAC.

Paul Bosco: so... how is W3C going to solve the codec problem and such? 1/2 ;-)

Steve Bratt: by Christmas ;-)
... summarizing how W3C works... a workshop is an early part of the process...
... in some cases, the outcome of a workshop is "no, there isn't a critical mass headed in the same direction that we can help with"
... but here I see lots of interest.
... Microsoft is engaged in other areas of W3C, and I'm confident we can talk with them.
... regarding Advertising, I know of a Mobile advertising agency... [missed some...]
... we have liaison with Open Mobile Alliance [OMA] ...
... the norm with workshops is that the workshop chairs draft a report, cirulate it to participants for review, and then W3C publishes it
... the report typically concludes with recommendations; anything from "no action" to "have another meeting" to "form a working group" etc.
... I heard about updates to existing specs: HTML, DOM, CSS, SVG, SMIL
... so I can imagine a group to represent requirements of this video community as input to those working groups
... a theme here was "making video a 1st class citizens" and we really care a lot about the architecture of the Web...
... what [makes it all work] is when you can freely link across ...
... re codecs... that's an area we intend to follow up on
... one of W3C's distinctive features is our technical staff. We have a fellows program, if any organizations here want to contribute people-time

<Chris Lilley> W3C Fellows program http://www.w3.org/Consortium/Recruitment/Fellows

Paul Bosco: how many people would be interested to meet in Q1 2008?

[at least, two thirds of the audience]

Many thanks to the participants, organizers, Chairs, and Cisco, as a host.

ADJOURN.