<cpn> scribenick: cpn
Chris: Before we start, any AOB items?
<Barbara_H> What is the charter status of the Media & Entertainment IG and Media Working group?
Chris: This is slightly unusual for
us, it's not a strictly media related topic,
... but given the new Web & Networks IG has been proposed,
there's enough overlap of interest for our members
... that we wanted to invite the leaders of the Web & Networks
IG to present, and discuss areas of common interest and
collaboration between our IGs.
<kaz> scribenick: kaz
Chris: (asks all if it's OK to record today's call and publish the audio publicly (like the minutes from this call), and there were no objections)
<cpn> (Slides)
<kaz> (Recorded video)
<cpn> scribenick: cpn
Eric: Thank you for the opportunity
to share the new, to be formed, Web and Networks Interest
Group.
... The charter just passed AC review, so should be formed within a
couple of weeks.
... I'll give some history of how this IG came to be formed. We had
a Web5G workshop in London last year,
... to explore the opportunity with 5G and the Web. We had a lot of
interesting technical presentations and discussion.
... At the end of the workshop, we asked ourselves what are the
problem statements that we're trying to address?
... We were not able to articulate succinctly what those problem
statements are.
... We also noticed that we did not have all the key stakeholders
in the room, in particular, the browser vendors.
... We decided not create a formal group, instead an informal task
force to explore and to find the problem statements that resonate
with the ecosystem,
... and then invest some upfront effort in socializing and outreach
to enrol the key stakeholders.
... The important thing was having all the right stakeholders agree
that these are problems they are interested in solving.
... Before any technical specs are developed, we want to make sure
there's a high probability of implementation when technical work is
done.
... (Slide 2) I'll describe the problem statements, potential
solution approaches, share some sample use cases, and then give an
overview of the mission and scope of the IG.
... I'll also explain how the Web & Networks IG will work with
other groups inside and outside of W3C.
... It's important not to duplicate work, also not confuse the
market and stakeholders as to what's happening.
... I'll give an update on the ecosystem enrolment and outreach
we've been doing.
... One of the purposes of this call is to encourage all of you to
join, when the IG is formed.
... (Slide 3) There are three problem statements.
... 1. Web browsers and apps today function agnostic of the network
types.
... There are some basic tools in Chrome for developers to simulate
how their apps might perform under different network types,
... but there's no capability of doing that during runtime.
... There are also a lot of capabilities at the protocol level, to
optimise media and video transmission.
... Web apps are moving from native to the browser. It would be
great to introduce new browser capabilities to give the deveoper
information, e.g,. on network quality and performance during
runtime.
... 2. This came from network operators: developers have limited
information coming from the browser to anticipate what kinds of
resources are needed for their app.
... There are techniques like DPI (deep packet inspection), but
there's a lot privacy concerns and the community has pushed back on
DPI.
... If there's a way for some of these requirements be communicated
between app and network, it would be useful.
... 3. App developers need better tools to test their applications
under different network conditions.
... For example, to compare how apps might run on the edge, vs on
the cloud.
... As the work evolved, the scope extended beyond 5G, so the IG
will explore how web apps will work on all kinds of networks: wifi,
4G, LTE, 5G.
... Ecosystem stakeholder enrolment and participation is key, so
we're focused on getting all the stakeholders to participate in
this work.
... (Slide 4) You'll be familiar with the waiting experience (video
buffering).
... This depends on various factors: the size of the upload, the
kind of network exchange, latency vs bandwidth tradeoff.
... Also, time for processing on the edge vs the cloud.
... Device capabilities, battery condition, processing power, all
come into play.
... (Slide 5) This is a high level approach to solutions. AT&T
were emphatic to communicate this with the community.
... We're looking at providing explicit hints in both directions
between web browsers and the network.
... The community has historically resisted techniques like DPI,
and invasion of privacy.
... These hints could be early hints or instantaneous hints. For
example, someone who is moving around and using the 5G network: the
cells are very small so there are a lot of handovers moving between
cells.
... So having instantaneous hints and the ability to adjust on the
fly would be very useful.
... That's the general direction we're proposing. We would
investigate this as well as other ideas with members of the
ecosystem.
... (Slide 6) To enhance the user QoE, we would introduce
contextual awareness between apps and the other components: the
device, network, browser, OS, cloud service or the CDNs.
... There are various aspects of contextual awareness. First, what
kind of application is it? Video, web browsing, what metrics to use
to measure the experience and adjust on the go?
... Another aspect is the network context: what's the condition of
the network at that moment, conjestion level, back haul, latency
and throughput, all need to be taken into account.
... Device context, what kind of device is running the app? And
user context: allowing the user to indicate their preference -
whether they want the cost of running a workload in the cloud or on
the edge or device.
... Also, Environmental context, relay services available at the
moment.
... (Slide 7) I'll share a sample use case. This is a machine
learning workload, running on the cloud vs on the edge.
... There all kinds of tradeoffs doing this kind of workload.
... My colleague Sudeep Divakaran, one of the IG co-chairs, ran
this workload.
... [A picture of a kingfisher run through the Berkeley cloud
service].
... (Slide 8) The processing time to get this result was quick, 0.1
seconds, but the round trip time is 3.4 seconds.
... The result is detailed and quick, but the waiting time of
sending the request to the cloud took 3.4 seconds.
... (Slide 9) Running it on the edge, it takes longer, 0.18
seconds, not a huge difference, but the round trip time is a lot
faster, about 7 milliseconds, but you compromise on the detail in
the results.
... This illustrates the choices the user can make, and the service
provider and network service provider.
... If you extend this use case to other workloads, there's a lot
of variation.
... Having the intelligence to communicate between the networks and
browser and apps would be useful.
... (Slide 10) IG Mission and scope. The mission is to explore
solutions for applications to leverage network capabilities.
... The goal is to achieve better performance and resource
allocation both on the device and networks.
... Scope includes hints, explicit hints, also finding ways to
expose specialised services and applications.
... For example, differentiative services, integrated services, 5G
network slicing, edge computing.
... In terms of applications, there's AR, VR, streaming, workload
balancing between edge and cloud for machine learning
inferencing.
... The third part of the scope is exposure of aggregated web
metrics, to allow you to monitor for troubleshooting purposes, and
make adjustments at runtime.
... (Slide 11) Coordination with W3C groups. We listed the most
relevant groups, but we also want to be inclusive, get inputs from
everyone and coordinate.
... Important to coordinate with M&E IG in particular, we don't
want to duplicate here, and we want to make sure members of this
group are participating in both IGs, to get the use cases and
requirements.
... As this is an IG, we won't do any specification development.
The work will be farmed out, e.g., to WebRTC WG, passing to them
the use cases and requirements developed in the IG.
... (Slide 12) Coordination with external groups. There are lots of
standards bodies working in related areas, e.g., 3GPP, MPEG.
... We'll ensure we're liaising with those groups, we want to stay
within W3C's core competency and not do work that's a core
competency in those other standards groups.
... (Slide 13) Regarding stakeholder enrolment, we have categorised
stakeholders: app developers, OS vendors, networks, broadcast
networks, content providers, semiconductors, etc.
... The logos show a sample of companies we've talked to in
workshop, the task force, or the meeting at TPAC last year, or one
on one conversations to get input.
... This effort is ongoing, so this gives an indication of the
companies we've been getting input and interest to
participate.
... In the AC review response, over 20 companies supported the
charter. Three members left comments: only two minor ones, and
we'll make changes and form the IG.
... It should be up and running in a couple of weeks.
... To end, I want to encourage everyone to look at the IG, and we
want to welcome you to participate in the discussion to gather use
cases and requirements, then figure out who to liaise with going
forward.
... The two co-chairs are Dan Druta from AT&T, and Sudeep
Divakaran from Intel. I will continue to be involved, on the
outreach.
<kaz> scribenick: kaz
Chris: Thank you for presenting
today!
... In the M&E IG we look at some network related aspects,
e.g., adaptive media streaming is an important technique for
managing media delivery over changing network conditions.
... We have some APIs in development now for providing information
to web applications to optimise streaming delivery, e.g., Media
Capabilities.
... How do you envisage the collaboration between the IGs, as you
get started?
Eric: As a first step, we'd ask for
contribution of use cases. It would be great to get key use cases
from M&E IG already generated, where you see relevance.
... We can look into them rather than starting from ground zero.
Also, having you part of the Interest Group, so we don't do
anything in conflict.
... Also defining some requirements to feed in to the
discussion.
... As we compile use cases and requirements, we're not conflicting
with each other's work.
Chris: I see our two groups as being
quite complementary, so we'd want to keep in close contact.
... Do you have any other use cases some of the stakeholders
mentioned?
Eric: Note that the IG isn't formed
yet. We have some from within Intel from our business units, we're
planning to bring those. From other stakeholders, many say this is
interesting work, things we're thinking about.
... Hopefully once the IG is formed, and the call for use case
contributions is made, they will also contribute to the
discussion.
Chris: How do you see the group would be organized, phone calls, emails and GitHub?
Eric: I would defer to Dan and Sudeep
on that.
... I imagine a combination of calls, emails, we're planning to be
at TPAC on the Tuesday.
... I would imagine before the F2F in Japan (TPAC 2019) we'll have
a number of conference calls to start gathering use cases, to make
the F2F productive.
Chris: Anyone else have observations or questions?
Glenn: Thank you for your
presentation.
... Regarding other organizations to have liaisons with, I'd
recommend consideration of DASH-IF and Streaming Video Alliance in
addition.
Eric: Sure, as we gather requirements, please raise that, so the Chairs can handle that and have the right liaisons.
Chris: That's an interesting case of the complementary interest between the IGs.
Eric: We recognize the importance of cross coordination with all of you.
Chris: Any more questions?
Eric: That's it! Thank you for the opportunity to share this with you.
Chris: Thank you for presenting today, it's really interesting, and I wish you success in getting the IG up and running.
Chris: We have a charter review open
right now, for the M&E IG, review is ongoing, please
respond.
... Feedback on the charter is welcome, or topics for these regular
calls.
Chris: We plan to return to looking at the media specs in incubation. June 4.
[adjourned]