Meeting minutes
WebXR-WebGPU-Binding#4 Wide Review for WebGPU and WGSL
<ada> https://
ada: if you have opinions, please weigh in
… the issue is in the webgpu repo
… we don't need to discuss here
… please take a look at the current spec and give feedback
… marker tracking
… the 2 people that want to implement it, the native API can only do image or qr tracking
… can't do both
… a short discussion could be useful
… I was thinking that if a user submits a marker, it's a lower barrier of implementation
… if the browser can say which ones it can track
… developers can ship a bunch of markers
… this has the downside that it only works on android
marker-tracking#10 Marker tracking, Status update?
<ada> https://
ada: there is an issue for this
<ada> current meeting agenda: https://
<ada> rik: it seems like QR tracking is probably a lower bar for entry
<ada> cabanier:
cabanier: marker tracking doesn't seem too useful. Too hard to set up
… qr code tracking seems like it fits the platform better. Very similar to planes tracking
bialpio: we specced out marker tracking as distinct from qr code tracking
… hololens allows to find qr code in the scene
… sites can find where the qr code is in space
… this is why it's difficult to come up with a spec
… because it's any qr code vs predefined images
… the other use case is a way to provide shared experiences
… if we know that they are looking at the same marker
… as for the status of the API, we have not looked into bringing it further
… it's behind a flag
… and you get a score
… we don't give any guarantee if an image will be tracked well
… there are no promises about quality of tracking
ada: shared spaces are important to me
… because it drops the need for shared anchors
Leonard: a qr code is just a special case of an image
ada: I think the issue is that android can't track qr codes
… so they don't have enough detail
Leonard: yes, so we have that problem
… I have code running on android that could handle QR codes
… there are lots of qr code formats
ada: so you're saying some qr codes could be tracked?
Leonard: yes
… for qr codes, you shouldn't assume that all input is good
ada: we shouldn't do open urls?
Leonard: maybe have a flag that it's verified
… add site restrictions
simon: we do a lot of image tracking
… the user flow is to open the camera and then the browser has the ux
… our use case for qr code is tiny
… image marker is the thing you should use
… we have our own markers that can easily be tracked
… maybe image anchors are separate
… if we have more camera access, we could do it
ada: this spec is more privacy preserving
bialpio: klaus experimented with pasting an image over a qr code
… so there might be some tricks we could do
… or how big of a burden it would be to the creato
Leonard: is this an attempt to have a custom artwork show up on printer material?
… or is it predefining on what is allowed?
… for instance, could a logo of a wine bottle work?
… a qr code would look out of place there
anchors#75 first pass at persistent anchors
<cabanier> https://
Wrong URL. Correction: https://
Rik: Get a unique ID for anchor based on URL that allows persistence
… created methods for CRUD.
… Pass on questions, comments, or concerns to Rik
Leonard: Anchors are not shared. Need to share data between platforms
[Note msome previous comments are from Rik]
Simon: Events for discovering anchors
Rik: Anchors can be restored. If doesn't exist, then process fails.
Rik: Does this capability need to be provided? Does a know list need to be made available?
Ada: Probably. Concerned about time delay between request and response.
Rik: Please file issue. Simon to create issue
bialpio: Are there any other systems that use an event based system for anchors
Rik: OpenXR has storage, but not returning lists
Rik, Simon, bialpio: Discussion about site vs. device for polling of anchors
Leonard: Is there an accessibility issue with being in "same space"
Rik: Oculus has space size of 8m^2. Android may be different
bialpio: Android would need cloud anchors. Process is much trickier
real-world-geometry#35 First proposal to add room capture
<ada> https://
Rik: With Oculus room capture does not always happen. This feature would allow the site to detect and instruct the user
bialpio: Potential for user confusion between room (boundary) capture and content capture
Simon: Different h/w approaches the problem differently and the applications reflect those differences
Rik: ARCore & ARKit (w/o lidar) both return no planes on startup
… may need a flag to provide developers what the initial state is
bialpio: Initial planes are not stable
… trying to determine what the user/developer experience would be
Simon: Use API to do a pre-warm-up?
biolpio: Doesn't really understand Oculus. ARCore has a start-up delay until a stable set emerges
EOM