14:05:13 <RRSAgent> RRSAgent has joined #element-capture
14:05:13 <RRSAgent> logging to https://www.w3.org/2022/09/14-element-capture-irc
14:05:15 <Zakim> Zakim has joined #element-capture
14:08:39 <dom> Meeting: Element Capture  - TPAC 2022 breakout
14:08:53 <dom> Chair: Elad_Alon
14:09:03 <dom> Agenda:  https://www.w3.org/events/meetings/ee306db5-9954-4674-a61d-ff68498b4608#agenda
14:09:11 <dom> RRSAgent, make log public
14:09:19 <dom> RRSAgent, this meeting spans midnight
14:17:08 <dom> RRSAgent, stay
14:17:10 <dom> Zakim, stay
14:17:10 <Zakim> I don't understand 'stay', dom
15:14:58 <dom> dom has joined #element-capture
15:52:36 <englishm> englishm has joined #element-capture
16:12:25 <dom> agenda+ breakout
18:15:49 <dom> dom has joined #element-capture
19:19:09 <jesup> jesup has joined #element-capture
19:19:25 <jesup> is there a zoom session for element capture?
19:19:37 <jesup> the link in the agenda led to an empty room
20:15:05 <dom> dom has joined #element-capture
20:30:10 <dom> dom has joined #element-capture
21:07:02 <dom__> dom__ has joined #element-capture
21:52:59 <dom> dom has joined #element-capture
21:58:45 <lgombos> lgombos has joined #element-capture
22:00:20 <mfoltzgoogle> mfoltzgoogle has joined #element-capture
22:00:25 <mfoltzgoogle> Present+ Mark_Foltz
22:00:44 <seukyoon-kang> seukyoon-kang has joined #element-capture
22:01:11 <mfoltzgoogle> Slides: https://docs.google.com/presentation/d/e/2PACX-1vSBwa1kvYKY14-nh3Wd523X7Nzy-fi-sBoplE4pBLGkOGBkTpR1u5hA5s2UDg4D1ICip6lYXA9bfKYl/pub?start=false&loop=false&delayms=3000
22:01:25 <seukyoon-kang> thank you!
22:01:54 <ericc> ericc has joined #element-capture
22:03:42 <mfoltzgoogle> zakim, start the meeting
22:03:42 <Zakim> RRSAgent, make logs Public
22:03:44 <Zakim> please title this meeting ("meeting: ..."), mfoltzgoogle
22:04:19 <mfoltzgoogle> meeting: Element Capture
22:04:35 <mfoltzgoogle> zakim, meeting: Element Capture
22:04:35 <Zakim> I don't understand 'meeting: Element Capture', mfoltzgoogle
22:05:32 <mfoltzgoogle> Scribenick: mfoltzgoogle
22:05:44 <jib> jib has joined #element-capture
22:06:27 <mfoltzgoogle> eladlon: Screenshot of me and Mark, docs can embed a videoconferencing app
22:06:44 <mfoltzgoogle> ...One of them captures a portion of the screen and transmits that remotely
22:07:23 <mfoltzgoogle> ...A couple of options, transfer stream to the video conferencing iframe
22:07:31 <mfoltzgoogle> rrsagent, generate minutes
22:07:31 <RRSAgent> I have made the request to generate https://www.w3.org/2022/09/14-element-capture-minutes.html mfoltzgoogle
22:07:50 <mfoltzgoogle> msw: In screen capture, for training videos, web app might capture a portion of the content area
22:07:55 <jib> Sorry I joined late. Is there a link to the slides?
22:08:08 <lyf> lyf has joined #element-capture
22:08:08 <mfoltzgoogle> jib: Slides: https://docs.google.com/presentation/d/e/2PACX-1vSBwa1kvYKY14-nh3Wd523X7Nzy-fi-sBoplE4pBLGkOGBkTpR1u5hA5s2UDg4D1ICip6lYXA9bfKYl/pub?start=false&loop=false&delayms=3000
22:08:41 <mfoltzgoogle> ericc: Would occluded portion of the elements being captured?  Danger of the user exposing information they cannot see
22:09:42 <mfoltzgoogle> eladalon: This can happen today, when users type input it can show up in the capture.
22:10:32 <mfoltzgoogle> <?>: In iOS, we have a mitigation that camera capture requires a preview
22:10:46 <jib> q+
22:11:21 <mfoltzgoogle> <?>: The concern is not malicious sites, but users knowing what they are sharing
22:12:18 <herre> herre has joined #element-capture
22:12:24 <mfoltzgoogle> <jib>: Built on getDisplayMedia but also getViewportMedia, very different security properties.  Which one?
22:12:38 <iclelland_> iclelland_ has joined #element-capture
22:12:49 <iclelland_> q?
22:12:56 <jib> ack jib
22:12:57 <mfoltzgoogle> q- jib
22:13:28 <mfoltzgoogle> eladalon: Requires for getViewportMedia, requirements were difficult to comply with
22:13:38 <mfoltzgoogle> jib: Non-compliance not a good argument
22:14:22 <arthursonzogni> arthursonzogni has joined #element-capture
22:14:24 <mfoltzgoogle> eladalon: See if we can relax from getViewportMedia, but if we can't, then stick with it.
22:14:56 <youenn> youenn has joined #element-capture
22:14:57 <mfoltzgoogle> youenn: Trying to understand, is the interest to reduce requirements from gVM, or to fix the occlusion issue
22:14:58 <fluffy> fluffy has joined #element-capture
22:15:22 <mfoltzgoogle> eladalon: Too early to stay, may not be possible, but it is an interesting topic to drive adoption
22:15:41 <mfoltzgoogle> rrsagent, generate minutes
22:15:41 <RRSAgent> I have made the request to generate https://www.w3.org/2022/09/14-element-capture-minutes.html mfoltzgoogle
22:16:04 <mfoltzgoogle> ... Two requirements for gVM: cross-origin isolation, and opt-in from document-policy
22:16:18 <mfoltzgoogle> ... Sounds like we could use it for element-level capture to reduce requirements
22:16:21 <mfoltzgoogle> q
22:16:46 <mfoltzgoogle> ... Document policy is sufficient, if others have ideas, lets discuss
22:17:56 <mfoltzgoogle> arthursonzogni: Seems sufficient from requirements
22:18:10 <mfoltzgoogle> ... Include iframes that does not include policy?
22:18:47 <mfoltzgoogle> iclelland_: With document-policy, I consent to be captured by this origin.  All embedded frames also must consent or we don't allow it.
22:18:57 <mfoltzgoogle> arthursonzogni: Is there a race when creating iframes?
22:19:42 <mfoltzgoogle> iclelland_: We have a mode, similar to CSP-embedded.  I require this policy on all frames I embed, if they don't comply, they don't load.
22:20:00 <mfoltzgoogle> ... Could be useful here, have to be careful with dynamic content.
22:20:02 <mfoltzgoogle> q
22:20:08 <jib> q+
22:20:21 <mfoltzgoogle> ack jib
22:20:47 <mfoltzgoogle> jib: Security requirements should be the same, iframe or div could be an element.  To me the security requirements are the same
22:20:58 <herre> q+
22:21:07 <mfoltzgoogle> ... Requires cross-origin isolation, which prevents loading non-consenting documents
22:21:26 <ken> ken has joined #element-capture
22:21:38 <masaki> masaki has joined #element-capture
22:22:12 <mfoltzgoogle> herre: Could be useful to talk about use cases enabled by more relaxed permissions.
22:22:24 <msw> msw has joined #element-capture
22:22:35 <mfoltzgoogle> ... Some small part could be captured without the entire document.
22:22:59 <mfoltzgoogle> youenn: Iframe could impact size of 3p content.  getVM seems like the right baseline.  Figure out if we need to decrease security.
22:23:01 <jib> ack jib
22:23:03 <mfoltzgoogle> ... Always a hole.
22:23:05 <bajones> bajones has joined #element-capture
22:23:17 <mfoltzgoogle> herre: Worth trying.
22:23:20 <mfoltzgoogle> ack herre
22:23:26 <jib> q+
22:24:34 <mfoltzgoogle> eladalon: Find a way to isolate, to prevent data leakage.  May need CSS to isolate rendering.  Worth looking for it.
22:25:06 <mfoltzgoogle> ... You've created content and you want to show ads.  Ads doesn't want to be captured.
22:26:11 <mfoltzgoogle> youenn: These are two different projects.  Two different specs.  If you want to have it soon, do the security later.
22:26:20 <mfoltzgoogle> ack jib
22:27:20 <mfoltzgoogle> jib: Maybe an unpopular comment.  Self-capture, at some point, creating a new render target for HTML.
22:27:56 <mfoltzgoogle> ... We've backed in to a powerful feature.  If an HTML app needs to capture video, then it's worth questioning whether the app can do it itself.
22:28:32 <mfoltzgoogle> eladalon: We have web developers asking for it. Google docs and meet, non-Google developers asking for it.
22:28:41 <youenn> youenn has joined #element-capture
22:30:02 <mfoltzgoogle> q
22:30:29 <fluffy> q+
22:30:42 <mfoltzgoogle> eladalon: Why shouldn't we do this?  If we separate/stage those discussions, how do we make it accessible to more users.
22:31:37 <mfoltzgoogle> ... Why would we not want to provide this API.
22:32:34 <mfoltzgoogle> jib: Can these applications do it with canvas?  If element capture requires a permission, would it meet the use case?
22:32:57 <ada> ada has joined #element-capture
22:33:17 <mfoltzgoogle> youenn: Occluded content, there is a big difference with getViewportMedia.  Will the user understand the difference between element capture and gVM?
22:34:24 <mfoltzgoogle> eladalon:  For most applications, there is no difference between gVM and element capture.  For malicious applications, the same issue.
22:34:34 <fluffy> q?
22:34:44 <mfoltzgoogle> youenn: The same prompt, but the user mitigation is different? From the user's point of view, inconsistent.
22:35:20 <mfoltzgoogle> youenn: If you scroll down, understood that it is not shared.  In fact it is not the same thing.
22:35:33 <mfoltzgoogle> jib: With this new API, there is no "flashing."
22:36:02 <mfoltzgoogle> eladalon: You broke the page apart, could piece them back together.  Machine could never see that.
22:36:43 <mfoltzgoogle> ack fluffy
22:37:29 <mfoltzgoogle> fluffy: Surprising that what is being captured is different from what the user is seeing.
22:38:09 <mfoltzgoogle> fluffy: Prompts don't solve.
22:38:37 <mfoltzgoogle> eladalon: Danger is limited, content is what origin already knows
22:40:25 <mfoltzgoogle> mfoltzgoogle: These issues have already existed with capture from canvas and capture from video.
22:41:01 <mfoltzgoogle> eladalon: Link purpling needs to stop when you start capture.  But I don't see it as unique to element capture.
22:41:41 <mfoltzgoogle> youenn: getVM is more like screen share, this is different, rendering in the background.
22:42:01 <mfoltzgoogle> eladalon: Some issues are unique, some are shared.
22:43:04 <mfoltzgoogle> jib: These concerns are elevated, with occlusion.  We already have offscreen canvas.  If I see things are flashing on my screen, if it's occluded there are no signs.
22:43:45 <mfoltzgoogle> eladalon: Assume that the user won't notice the flashing if it's a single pixel.
22:44:28 <mfoltzgoogle> eladalon: We should mitigate concerns and make them happen.  Some tricks that the web page might do, we can do mitigations.
22:44:38 <mfoltzgoogle> youenn: If we can't see the content, then no mitigations.
22:45:07 <mfoltzgoogle> ... With occluded content, I don't see what we can do.
22:45:30 <ada> q+
22:45:43 <mfoltzgoogle> youenn: Iframe that is fully isolated?
22:46:10 <mfoltzgoogle> iclelland_: Fenced frames
22:46:46 <mfoltzgoogle> eladalon: Could be that size leaks information.
22:46:56 <mfoltzgoogle> ...about embedded content.
22:47:03 <mfoltzgoogle> ack ada
22:47:41 <mfoltzgoogle> ada: Chair of IWMG.  One of the most requested features, is an API to view DOM scenes.
22:47:51 <mfoltzgoogle> ... Problem trying to solve for 5 years.
22:48:09 <mfoltzgoogle> ... Either too dangerous, or restrictions made it unusable.
22:48:25 <mfoltzgoogle> ... Play youtube in WebXR.
22:48:43 <mfoltzgoogle> ... Might have some related places where we encountered problems.
22:49:06 <mfoltzgoogle> ... DOM layout that is not readable by the user.
22:49:36 <mfoltzgoogle> ... Adding on the ability to a fake interaction on the projected view, redirect back to the original DOM without synthesized clicks?
22:51:22 <mfoltzgoogle> eladalon: Seems orthogonal to me.
22:51:41 <mfoltzgoogle> jib: Out of scope for screen capture spec.  End up tripping up malicious use cases, scams.
22:51:48 <mfoltzgoogle> q
22:52:09 <mfoltzgoogle> yoenn: Do you need user specific rendering? Being rendered based on user specific information.
22:52:24 <mfoltzgoogle> ... Any browser instance would produce the same information.
22:52:48 <mfoltzgoogle> eladalon: Mostly personalized content.
22:53:24 <mfoltzgoogle> youenn: Maybe you want an HTML to video converter that doesn't depend on the user.
22:54:33 <mfoltzgoogle> eladalon: Could turn off link purpling, autofill, etc. for the user, accessibility settings
22:55:36 <mfoltzgoogle> youenn: Accessibility should be protected
22:56:08 <mfoltzgoogle> herre: Users need informed consent.
22:56:32 <mfoltzgoogle> jib: Browser to steer users towards better choices than sharing the full screen.
22:57:10 <mfoltzgoogle> eladalon: Google feedback tool uses gDM, we could do better.
22:57:51 <mfoltzgoogle> jib: Working on better choices, windows, tabs.  Not sure this is moving in the right direction.  Push back on the use case.
22:58:21 <mfoltzgoogle> ... Better ways to solve overlapping menus.  Smart cropping to an element.  Move the menu to the side.
22:59:06 <mfoltzgoogle> ... You can say crop to this element, and crop to coordinates, if you don't want to overlap.
22:59:27 <fluffy> q+
22:59:31 <mfoltzgoogle> eladalon:  Give applications a consistent way to remove private content, like chat notifications
23:00:21 <mfoltzgoogle> ... Giving the proper weights to the issues.  The occluding content that is private, more of a problem than getVM
23:00:44 <mfoltzgoogle> fluffy: Can you turn off notifications in browser apps, we are talking about in-content notifications.
23:00:55 <mfoltzgoogle> eladalon: ChromeOS does that
23:01:07 <mfoltzgoogle> ericc: ScreenCaptureKit does that.
23:01:08 <fluffy> ack fluffy
23:01:35 <mfoltzgoogle> eladalon: We deal with tradeoffs, current state is not great.  We could be allowing better applications to be built.  The perfect be the enemy of the better.
23:01:40 <ericc> present+
23:01:54 <mfoltzgoogle> rrsagent generate minutes
23:02:05 <mfoltzgoogle> rrsagent, generate minutes
23:02:05 <RRSAgent> I have made the request to generate https://www.w3.org/2022/09/14-element-capture-minutes.html mfoltzgoogle
23:22:25 <dom> dom has joined #element-capture
00:08:15 <eladalon> eladalon has joined #element-capture
16:49:40 <msw> msw has joined #element-capture
21:14:55 <dom> dom has joined #element-capture
22:00:27 <dom> RRSAgent, bye
22:00:27 <RRSAgent> I see no action items