Multicast + Second Screen�Discussion Questions

Jake Holland

2021-10-25

Outline

• Multicast CG: Mission & Motivation
• Discussion Questions
• Privacy Threat Model
• Incognito Mode?
• Anything else to think about?

Multicast CG

Mission:

to enable multicast IP transport for web traffic to efficiently solve scalability problems in networking and web operations.

https://w3c.github.io/multicast-cg/multicast-cg-charter.html

Why: User Experience (Effects of Congestion)

4

Observed goodput to large ISP by Time of Day (normal-traffic day, 100KB+ objects)

Why: User Experience (Effects of heavy Congestion)

5

Observed goodput to same ISP* by Time of Day (high-traffic day, 100KB+ objects)

* NB: effect present in most ISPs, but not always this clear a signal.��By eye: ~½ the goodput for a 9-hour peak

Key Problem Solved: Access Network Congestion

6

Cable Network Diagram By Saub09 at English Wikibooks, CC BY-SA 2.5, https://commons.wikimedia.org/w/index.php?curid=61793561

Deepest Useful Caches

Problematic

Congestion

7

Broadcast link capabilities can be leveraged by multicast? (up to?)

​

• Fiber (GPON, etc): yes (~3k/ONT)
• Cable: yes (~2k/service group)
• DSL: depends (~1.5k/chassis)
• PPP-based deployments can’t use broadcast
• Helps uplink bandwidth, but similar power usage
• Ethernet: usually (~2k in enterprise/university/apartment networks)
• Needs L2 snooping & replication capability--usually there, not always
• 3G & 4G: sort-of (with eMBMS: ~3k/tower, special signaling)
• 5G: yes (with Xcast: ~3k/tower?, normal signaling?)
• ATSC: maybe one day (~10-100k/antenna, will need special signaling)

​

(* Wifi in homes may need updates--solutions exist, deployment spotty)

Access Technologies: gain estimates at bottleneck links

• Climate Impact
• Internet=3.7%* of global carbon footprint (2019 estimate)
• 1% is from watching video (300m tonnes)
• 24m tonnes from video games (chiefly downloads?)

​

• Cost of delivery & services
• Network capital costs driven by peak load
• Power needs/provider costs scale with traffic volume
• Lower costs + competition => lower price for users

8

Other Effects

* “Why your internet habits are not as clean as you think”, 2020-03-05, BBC

Why: Avoidable Traffic (game/os downloads - new releases)

9

Under 100 streams: >40% reduction in peak load to ISP (high-traffic day)

Why: Avoidable Traffic (web video)

10

1 stream, >15% reduction in peak load to ISP (popular sport event day)

Privacy Issues

• Next-hop join exposure to LAN is fundamentally different from TLS/unicast
• IGMP/MLD MUST be visible at next-hop (e.g. wifi router), typically visible to all on LAN
• Multicast Group can probably be tied to content for scalable content
• Partially addressable by other means? (e.g. random mac?)
• Upstream it’s aggregated, but “in-household” similarly exposed to ISP
• Malicious site fingerprinting footprint increase?
• Does it matter if e.g. malicious ads can leverage this to expose something to LAN?
• Does it matter that e.g. reference to xx.local already exposes such info?
• (is this whataboutism?)

Privacy Threat Model?

https://www.w3.org/TR/remote-playback/#personally-identifiable-information

• Any observed attacks/abuses on this? (Any monitoring to look for it?)
• Any risks considered for exposing web pages to compromised local devices?
• Or to a hostile network operator?

Thoughts on Incognito mode?

• Any consideration of whether Incognito mode helps here?
• Should discovery be disabled when trying to avoid leaving more fingerprints?
• Necessary? Sufficient?

​