[e2e] More on bandwidth charging

[Jon Crowcroft]

the problem with these approaches is that users adapt - they adapt
their behaviour, their application behaviour and their workload mix -
for example, when going from 56kbps to xdsl of cable modem, MOST
People eventually learn that keeping all your telecommute work files
at the workplace and doing remote samba/nfs and X or citrix or nt
console is far more long term effective, but it also switches MOST
people from being in the lower category to the higher

when you go to higher bandwidth still, in our local uni experience
(which is isually a predictor of what happens to domestic use 5-10
yerars later) people start to run low latency continuous media (games,
hi fi audio) all the time

so all these experiences based on early experience in commercial
settings (esp. ones extrapoolated from simpley movign the curve out)
are imho on the VERY low estiamte side in the end


In message <3AE5EE93.BA5A1921 at cs.columbia.edu>, "Henning G. Schulzrinne" typed:

 >>From isen.com:
 >>
 >>Ken Poulton [poulton at labs.agilent.com] a Palo Alto CA fiber 
 >>activist, responds to my query in SMART Letter #52 about whether 
 >>1 Gbyte per month is a reasonable throughput budget:
 >>
 >>  "My e-mail record on my at-work system says: 
 >>   0.04 GBytes/month last year in personally-handled mail 
 >>   0.02 GBytes/month in unique weather data messages 
 >>   (sent hourly to Bay Area windsurfers) 
 >>   1.50 GBytes/month when you count the 400 weather 
 >>   subscribers
 >>
 >>  "My home Unix system (used as a remote work terminal): 
 >>   1.9 GBytes/month
 >>
 >>  "So even these rather heavy corner-case uses are only a 
 >>   few GBytes per month.  Most users would be hard-
 >>   pressed to get to a GByte/month until they turned on 
 >>   streaming video or Internet radio stations. Those 
 >>   could get your usage up to around 10-20 GByte/month at 
 >>   24 hours/day.
 >>
 >>  "In the Palo Alto Fiber to the Home Trial, we 
 >>   considered a service that would be limited to 1 
 >>   GByte/month.  I continue to believe this is a 
 >>   perfectly reasonable way to operate, but I can tell 
 >>   you from experience that it's a hard sell to consumers 
 >>   when you're offering a 100 Mbit/s pipe that can (in 
 >>   principle) use up your month's quota in 80 seconds. 
 >>
 >>  "For starters, we settled on an unlimited service that 
 >>   is 100 Mbit/s to the hub, but uses a shared 10 Mbit/s 
 >>   connection to the Internet.  Locally-served services, 
 >>   such as video-on-demand, can make good use of the 
 >>   local 100 Mbit/s, but Internet access (where the BW 
 >>   gets expensive) will be limited. We have not turned on 
 >>   the system yet, so we can't tell how it will work out 
 >>   yet. 
 >>
 >>  "The problem is non-trivial: $100/month only pays for a 
 >>   *continuous* usage of about 100 kbit/s.  You can get a 
 >>   *lot* of benefit from peak rates of 100 Mbit/s, but 
 >>   you have to deal with the people who will suck you dry 
 >>   with a server farm if you let them. 
 >>
 >>  "A pay-per-GByte system (like 1 GByte free, $100/GByte 
 >>   thereafter) will never ding most people, and charge 
 >>   big users appropriately. But it makes consumers 
 >>   (justifiably) nervous. What if you get a $1300 bill 
 >>   when your son leaves on the Internet radio 24 hours a 
 >>   day for a month? Simply cutting people off doesn't 
 >>   work either.
 >>
 >>  "What I think is needed is to allow consumers to choose 
 >>   full speed access up to 1 GB/month, and then throttled 
 >>   BW after that. Your connection may get slow, but 
 >>   doesn't die.  And you don't get a huge bill by 
 >>   surprise.  I suspect this will need special support in 
 >>   the routers or some kind of special billing box. 
 >>   Anyone care to support this in their routers?"
 >>-- 
 >>Henning Schulzrinne   http://www.cs.columbia.edu/~hgs

 cheers

   jon