Meeting minutes
Sudeep: our focus will be on edge computing today given the current momentum behind this topic
… our goal is to identify next steps, in particular towards TPAC
… we'll also talk about the upcoming TPAC 2022
… and a summary of our existing & new topics in the IG
Edge Computing Workstream
Edge Computing Exploration github repo
Client-Edge-Cloud coordination Use Cases and Requirements
Max: I've updated the document based on the comments from the last meeting
… I'll review these updates
… and Michael sent a PR earlier today that we can also discuss
… We've identified editors (separately from contributors)
… the co-editors are Michael, Song and myself
… There was also a request to highlight the benefits of offloading to the cloud or edge in the use cases
… which we've done through use cases - e.g. in the Machine Learning use cases involve really intensive computing which depending on the underlying hardware may imply very different outcomes
… e.g. whether they have a GPU or not
Michael: maybe a section summarizing the benefits would be useful
Max: +1
Max: another comment was a suggestion to refine the gap analysis
… it now emphasizes what the current specs cannot support
Michael: I think the bullet points in section 5.2 may deserve to be a section standing on its own
Max: let's iterate on this indeed
Max: another update is about the the conclusion & way forward section (8.3)
… we list 2 options: divide the work across existing relevant Working Groups,
… or we establish a dedicated Working Group for this problem
… the 2nd option helps with a unified architecture - we can coordinate with other groups as needed
… if we agree this is the right approach, we could proceed with writing a charter
Michael: I think we need to list the potential deliverables
… this would help establish which may need a new Working Group
… it's quite likely this would be needed, but there will also be a need to coordinate with other groups
Sudeep: we also need to identify the stakeholders who would need to support the work & a potential charter
<McCool> (since I interrupted so much, I will be quiet for a while ;)
cpn: I think we still need to improve the understanding of requirements before proposing a WG
… the exploratory work may not be a good fit for a WG
Dom: Agree, CGs are more a place for the exploration
… Overall I think mapping proposed solutions to new specs or changes to existing specs, protocols, APIs, there's room for pre-standardisation work
max: I agree with your points - this needs further work before proposing a WG
… a CG sounds like an interesting option
Song: +1 on iterating in the IG or in a CG to draw a clearer picture of what we need to build
… having a single group would help making progress rather than split across groups
Michael: +1 on more incubation; I think the document itself also needs more iteration
… getting more and broader input on it would be good
… a W3C workshop might be a good way to achieve this
… a CG would be one way to do so but it has downsides, but an IG is already a good fit
Dom: I agree having a single group would be more comfortable
… When we get to the right stage for standardisation, I expect we'd need to split it across multiple groups
… e.g., WASM work going to that WG
… We should look at it as a collection of useful pieces, rather than something that must be adopted all together
… An option 1 approach may be more likely to get to where we want
… On getting broader input, the first thing to do would be to publish as an IG note, which signals the group wants input on it
… When the PR is merged, are we ready to get broader attention on the document?
… For TPAC, I was thinking more of an informal workshop, could lead to a formal workshop later
… Use TPAC as a way to gather presentations on the topic, prerecorded, different perspectives, to add to our thinking
… Relevant to how edge computes intersects with the web. A formal workshop probably needs a bigger sense from the community
Michael: I think the document needs a bit more work before being published as an IG note - but I think we could there by the time of TPAC
sudeep: +1 on gathering input from beyond the group
… a mini-workshop or a breakout session at TPAC might help with that
… this sounds like pre-requisites before a WG
max: I can look into creating a CG if there is support
sudeep: there will be demos from Max, Louay around edge (not necessarily edge offload)
… maybe we should look at organizing a formal workshop before the end of our charter
… and use TPAC, this document as drivers toward that
Piers: any input from existing companies that provide edge services such as cloudflare, fastly
… (beyond Alibaba that presumably has some as well)
Max: Alibaba has an edge service, but not an implementation of the proposed solution which would need more standardization support
Dom: I hope we could attract those companies to a workshop
… I agree they're critical to making this successful, we need at least two
Michael: +1 - we need to plan on how to attract these companies which have solutions pretty similar to what is being discussed
… we should make sure the stakeholders listed in the business categories need to be well-represented
Louay: our demo is focused on edge in the context of 5G
… we showed it at the Media Web Symposium - I plan to record it as a demo
Dom: Breakouts are more limited this year, 1 hour for broad visiblilty, but probably not 2 hours
… Could have a longer timeslot for the TPAC meeting, that could be used for demos, or discussion
… The number of people would be more limited, but would give more leeway for scheduling
TPAC planning
Sudeep: WNIG meeting on Sep 13, then breakouts on Sep 14
Michael: afternoon slots won't work well for remote participation to breakouts
Sudeep: maybe that points toward doing the mini-workshop post TPAC
Dom: The last breakout slot at 4:30 could be possible for people in Asia?
Sudeep: our agenda at TPAC will be structure around updates to our workstreams, with plenty of time for open discussion (with room for a possible guest speaker)
… it'll be posted on the wiki
Updates on other topics
Piers: the work on monitoring in IETF is on hold, but there is work happening in CTA WAVE around this
Chris: I'm talking with that group in the context of the Media & Entertainment IG - let's approach them together
<McCool> (sorry, I have to go, another call with SDW...)
Sudeep: not much progress on link performance prediction; haven't seen much progress on priority control either
… on Edge Computing, good progress on our offloading; not seeing lots of progress in intersection between MEC & Web
… Not much anew on network emulation either
… New topics have emerged recently: exposing network slicing (which are now exposed e.g. in Android)
… We also received feedback from the Games CG on the network APIs gap e.g. for download progress information
Dom: the Winter CG is looking at making browser APIs into non-browser runtimes such as node, deno, cloudflare worker
… possibly interesting overlap with our offloading explorations
Sudeep: also potential interest around multicast receiver for Web using edge nodes
<dom> i/<dom> Sudeep: our focus/Scribe+ dom