Meeting minutes
<Max> ?
Max: first part is the use case discussion
… will give a brief intro
… there was a related breakout session today
… about generative AI and Web
McCool: this is more about requirments for AI models
Max: sometimes the latency can be sensetive
… we wont discuss solutions in this session, more on the use cases
Song: about cloud game, I presented the game use cases on the web Network IG
… why I bring this case here is that, we have some web-based cloud game now
… with the commercial round of the cloud game, with 2 million users
… the biggest consuming part is the keyboard events
… to be translated
… from end to end time we can see the translation will be the major part of the how process
… we still talk about the ideal mechanism for that
… we came up with 2 solutions
… since the Web based solution is on top of the TCP IP network
… hard to get better perf on that
… so we ask if we can have @@1 network, with higher priority
… 2 solutions, 1 is to have MOQ, to be standardized in 3GPP, which will change the low latency situation
… that one is a long term evolution
… the 2nd one is the Cloud Edge Client Coordination
… we had another case for the Paris Olympic game
… we set up DanDruta
[Song is showing a game case of Black Myth]
… the web based rending effect is good in this game case
… the end ot end latenty is acceptable
McCool: how far are the servers and client?
Song: the edge is delpyed in each city
… at least in east part of China
[Song showing the architecture of the game]
… within 100km
McCool: reason this gourp is called Cloud Edge Client Coordination
… is that there is interaction between Cloud Edge and Client
… local network edge computing can also be done
… next use case, a home system for smart house
… people can give command by voice
… a home server can do AI
… one thing I try is can we have a web paltfrom in distributed fashion to handle distributed wokers
… not much standards around this
… cool things can be done, like people recognization
… this kind of thing can also be provided by 3th party
… can be a new business model
… or for security as well, as intruder detection
Max: for this use case, all the computing is happening on edge?
McCool: yes
Max: anything on the cloud to control it?
McCool: nop
… I can run the network down
… like a local network
@@3: in same domain, easy to solve the trust issue
… media server is easy to discover
McCool: HP is local, which is nice
@@3: what is the split between the work
McCool: either breaking down the pieces or sandboxing
Max: we can discuss the details in the CG meeting
… we have a document, a W3C group note for the use cases
<xfq> https://
Max: let's move to the demo part, provided by Alibaba Could
… the demo is to do the website acceleration
[Max showing the demo]
McCool: purely functional, server-less?
Max: yes
Max: we have done the first 2 parts
… the 3rd part is to discuss the gap and how we can prioritize the work
DanDruta: how does the server determin that there is an offload opportunity for edge?
Max: this is triggered by the user
… it's called central cloud control
… the central controls the cloud resources
McCool: this is a different model
… the appraoch I take, is move the client app and @@4
… how something is up and run is independent to the service
DanDruta: where W3C needs to be involved?
McCool: web workers let you offload the work
… what if the webwork can run different machines
Max: like what in the game use case, the cloud and edge can do the rending, the user side can receive the rended content
McCool: I might discover a computing resouce nearby when I try to run an demanding app
… for game use case, maybe I can download the video and offload the other pieces
DanDruta: the big question is who has relationship with who
… and which I do I trust and the business model for it
McCool: there are gaps, how to discover the service, how to trust them, how to do communicaiton, etc
Zolkis: what is mission in the space
… s/mission in the space/is missing in the space?
… can we use the existing things like Web bluetooth?
… we need to verify what we actually want to achieve
McCool: I think we are talking browser APIs here
Max: the edge run time is related to winter CG
… we need some cooperation with them
McCool: need to work on things like what kind of library is available
Max: 5 min left, we can do through the gaps and see if we have consensus for them
… this can be the base to start the CG work
McCool: we need to identity which gap we need to focus on
Max: the plan is to launch the CG after the CG
… and then start the incubation work in the CG
… we will not go the gaps one by one
Jeffery: 2 different kinds of use cases, one is the servers know each other
… how to combine them
<kenji_baheux> this reminds me of https://
<Zakim> jyasskin, you wanted to ask about commonalities between the client-controlled and server-controlled edge models
McCool: we have multi-stakeholders in the CG, with different prospectives
… e.g. China Mobile can run a service on their cloud tower, it can be chosen by the client
… you dont have to tell the app what to do, it can scan for the resources available
… dont need to tell the app how to do it
McCool: the CG will be launched soon, welcome to join
Max: thank you all for the discussion!