Policy, Etiquette and Manners in the Unlicensed Wireless Bands

1
Policy, Etiquette and Manners in the Unlicensed
Wireless Bands

• Eric Friedman Cornell
• Supported by NSF ITR-0205431

2
Collaborators

• Rutgers Wireless Networking Laboratory (WINLAB)
• Narayan Mandayam, Dipankar Raychaudhuri, Chris
  Rose, Predrag Spasojevic, Roy Yates
• Quello Center for Telecommunication Management
  Law
• Johannes Bauer, Steve Wildman

3
Outline

• Intro to wireless networking and the unlicensed
  bands
• Macro-issues in the unlicensed bands
• Market structure, legal and regulatory issues
• Micro-issues etiquette and manners
• Lessons from the Internet
• Congestion control and TCP-friendlyness
• Manners for Information
• Golden Rules, Fair-queuing and Serial Cost
  Sharing
• Manners, Pricing and Reputations
• Concluding Comments

4
Wireless Spectrum

• 2.4 (ISM) and 5 GHz (U-NII)
• Large bands for unlicensed devices
• Power and antenna restrictions
• Short range 50-300 feet.
• Compare to licensed bands
• Cellular phones
• TV and Radio

5
A Ubiquitous Zoo/Ecosystem?

• Licensed band are homogeneous, but not unlicensed
  bands
• Wireless networks
• 802.11
• Ad-hoc networks
• Wireless devices
• Keyboards, mice etc.
• Sensor networks
• Medical
• Household
• Micro-Electro-Mechanical-Systems
• Innovations

6
My Background

• Noncooperative protocols for the Internet
• congestion control, routing, webserving
• Cooperation, reputation and the rule of law
• Online such as ebay
• Designing online environments to promote trust
  and reputations (PACT project)
• In emerging markets
• Wireless issues
• Pricing and coordination
• Regulation and licensing
• Pricing network usage to promote social welfare
• Cornell Office of Information Technology usage
  based pricing

7
Legal Issues

• FCC has regulatory authority
• Complicating Issues
• Property rights
• University Policy forbidding interference with
  its wireless network
• Tribal lands
• Tricky question can I set up a network outside
  of Starbucks and undercut them?
• Starbucks can fix this with wire mesh, but its
  ugly.
• Trickier can I set up inside?

8
Underlying Goals

• Facilitate coexistence
• Encourage Infrastructure investment
• Provide safe harbor of some sort?
• Avoid wasteful strategic behavior
• Squatters
• Interferers
• Provide incentives for innovation
• New killer apps?
• Avoid too much lock-in

9
Licensing Options

• Note property rights should include
  frequency/power and location
• Permanent licenses
• Auctioned off perhaps
• Resalable?
• Short term licenses
• Real time auctions possible over the WWW
• Homesteading
• Related to safe harbor
• Land based
• No licensing

10
(No Transcript)
11
Revenue

• FCC collected billions in spectrum auctions
• How much are these bands worth?
• Technically, these could serve the public good
• But,
• 100million households 10/yr 1B/year
• How to collect the money?
• Tax on devices
• Problem is there is not much value
• yet

12
Land based rights

• Seems simple and logical
• Universities can prevent leakage onto their
  campus
• Starbucks can keep rogue wireless out
• Without wire mesh
• Some hold up issues
• Issues with public spaces
• But could be left to local government like
  zoning.
• Not been discussed much
• More emphasis on fancy things
• (Note Can use directional antennas to avoid
  borders)

13
Remarks on Licensing

• Each licensing design has s and s
• Each would encourage different types of uses and
  different degrees of innovation.
• Perhaps we should strive for diversity
• Also, simplicity seems important
• For example Repeated localized auctions seem far
  to unwieldy.

14
Regulating Unlicensed Bands

• If we have unlicensed bands then we do need some
  regulations.
• To avoid problems such as
• Using 802.11 to trash a bluetooth competitor
• Implementing a large infrastructure only to have
  it become unusable
• 802.11 in an apartment
• Used to be a significant problem with portable
  phones
• Note, cell phones are centrally controlled!
• General overuse (tragedy of the commons)

15
The Commons

• Benkler 2003, Benkler and Lessig 1998, Buck 2002
• Commons is unlimited more grass for more sheep!
• Perhaps true theoretically
• Is this reasonable?
• Requires incredible coordination, computation
  etc.
• Probably not a good model for the ubiquitous zoo
• Is it relevant?
• If it turns out to be true, then wireless will
  solve itself, so why worry
• Although, perhaps all FCC licenses should be
  finite time, just in case

16
Demand

• Is there enough spectrum for everybody?
• DOS had a maximum of 512k of memory
• Expected to be more than anyone would ever need!
• Why doesnt standard software run faster today
  than 5 years ago?
• Why isnt the Internet overloaded with the demand
  for broadband?
• Answer Demand is endogeneous.
• if you build it they will come
• Prediction however much there is, thats what
  well need.

17
Regulating the Unlicensed Bands

• I claim that this is indeed necessary.
• Similar to Buck 2002, Benkler 2003
• Although I have far less faith in cooperative
  decision making by groups because of
• Easy entry
• Very low costs, so even small transaction costs
  would be too high
• Endogenous demand and overuse

18
Lessons from the Internet

• We have a very good analogue the Internet
• Its a commons to a large extent
• Why does the Internet work so well today?
• (Aside from lots of black fiber these days)
• Congestion control (Floyd 2000)
• What might keep it working in the future
• Fair queuing and generalized cost sharing
  (Demers, Keshav and Shenker 1990)

19
Congestion Control on the Internet

• Mid to late1980s
• LBL to Berkeley (400yds) went from 32Kbps to
  40bps
• Problem TCP would resend dropped packets
• Leading to more congestion and more dropped
  packets!
• Solution congestion control (Jacobson 1988)
• Slow down transmission rate when you see
  congestion
• Works amazingly well and keeps the Internet
  running
• Even helps other things work
• Outcomes are fair
• Reduces losses due to selfish routing (Friedman,
  2002)

20
How Congestion Control Works

• Additive increase, multiplicative decrease
• Slowly increase data rate as long as no problems
  arise
• Reduce rate by ½ in case of a problem.
• Outcomes are fair
• Sort of

21
Equation-Based Congestion Control

• Should we require AIMD power control for
  wireless?
• No, even on the Internet it doesnt work
• Streaming media
• Alternative idea require all sources to be
  TCP-friendly
• Use the same amount of resources as TCP would
• Any method of achieving this is ok.
• Formula transmission rateRa/L1/2, where L is
  the average loss rate

22
Etiquette vs. Manners

• Etiquette using a spoon for soup
• Manners making your guest feel comfortable
• (paraphrased from J. Martin)
• TCP congestion control is etiquette, equation
  based rate control is manners!
• Etiquette Listen-before-talk or rules about how
  to implement spread spectrum
• None on U-NII, since many felt they would be too
  limiting
• Lots on ISM bands.
• Manners
• Inverse golden rule dont harm anybody worse
  than they are harming you. (Shenker 1995)

23
Fair Queuing and Fair-Share

• Focal question What equation should we use for
  wireless?
• Yet another Internet analogy queuing protocols
• Sources choose a data rate Ri and receive a delay
  based from the network.
• There are queuing constraints on the possible
  allocations of delays (Coffman and Mitriani 1980)
• But, the queuing protocal sets the actual delay
  functions.
• Examples
• FIFO
• Fair-Share/Fair Queuing

24
Fair Queuing and Fair-Share

• Fair-share is
• Fair every source gets at least their fair
  share of the network (Shenker 1995, Friedman and
  Moulin 1998)
• Strategically stable in a wide variety of
  settings (Shenker 1995 Friedman and Shenker
  1998 Greenwald, Friedman and Shenker 1999
  Friedman et. al 2002)
• Demand Monotonic (Moulin and Shenker 1990
  Friedman and Moulin 2000)
• Can do something similar for wireless?

25
Wireless Network Information Theory

• m-user multiple access channel
• (Formulae for m2, generalizes easily)
• E.g., several WiFi devices sharing a base station
• Shared (additive) Gaussian Channel
• Ri transmission rate of device i, Pi power of
  device i, N noise
• Ri ? C(Pi), R1R2 ? C(P1P2) (Cover 1975)
• C(x)log(1x/N)/2channel capacity
• Invert these to get D(Ri)? Pi, D(R1R2)? P1P2
• This is structurally identical to that for the
  Internet!
• Increasing R increases utility while increasing P
  decreases it.

26
Feasible Region
Required Power
P1P2 D(R1R2)
D(R2)
D(R1)
27
Relation to Fair-Share

• We could allocate information rates using the
  formula for fair-share.
• Fair-Share R1ltR2 then P1 D(2R1)/2 and P2
  D(R1R2)-P1
• Note it is natural to order devices by data rate
• Note scale invariance (leading to Aumann-Shapley
  pricing) does not seem natural
• Value
• Provides basic performance guarantees fairness
• Provides dynamic (and game theoretic) stability.
• Second best (1st best is not achievable)

28
Open Issues

• Note this is very preliminary!
• Computing the allocations in real time
• Practical constraints
• Does this generalize to include locational issues
  (such as shadowing)?
• Generalizations of Fair-Share exist which have
  the same normative and strategic properties
  (Friedman and Moulin 1999, Friedman 2002)
• Informational efficiency monotonicity (Benkler
  2003)

29
Foundations for Pricing and Reputations

• How do you get devices to provide services for
  other devices?
• Such as ad-hoc networks, peer-to-peer file
  sharing etc.
• Micro-Pricing (Mandayam in progress)
• Need micro-money and rule of law (Johnson et.
  al. 2000)
• Barter
• Need long term reputations
• What do you do with new devices
• Paying your dues equilibria (Friedman and Resnick
  2000)

30
Manners for Pricing and Reputations

• Many complex strategic issues
• Devices that intentionally harm competitors
• Devices that are intentionally fragile
• Free riding
• Social norms and the folk theorem.
• Are there a set of manners that work?
• Can you use a golden rule?

31
Concluding Comments

• Licensing
• This is a complex issue with important
  implications
• Must be based on realistic technological
  assessments
• Demand is probably endogenous
• Strive for Diversity and flexibility
• Manners
• There need to be some requirements on sharing and
  good behavior, but these cant be too specific.
• My proposals are preliminary
• Regulate manners not etiquette
• Overall
• Technology and policy are inextricably
  intertwined in IT regulation and policy.