The Real Congressional Agenda? (was Re: The House of Representatives on campus downloading)

[David Farber <dave () farber net>]

Begin forwarded message:

From: Brett Glass <brett () lariat net>
Date: May 6, 2007 11:50:18 PM EDT
To: dave () farber net, ip () v2 listbox com
Cc: smb () cs columbia edu
Subject: Re: The Real Congressional Agenda? (was Re: [IP] The House  
of Representatives on campus downloading)

At 06:44 PM 5/6/2007, Steven Bellovin wrote:

> I think you need to distinguish between bandwidth consumption -- or
> overconsumption -- and illegality.  While I agree that most of the
> bandwidth use you report is probably the result of illegal
> file-sharing, there are many legitimate uses for the same technology.

All burglary tools have occasional legal uses. ;-)

> To me, then, your note poses several questions.
>
>        Does the Internet need deployed QoS?  (There's been a lot
>        of work on general QoS, but this is a special case.)
>
>        Is there an inherent problem with wireless, or at least
>        with the type of wireless you use?

For some reason, several people on the list -- perhaps because
they don't understand the problem -- have asked this same question.
The answer is, emphatically, "no." If all of our users had wired
links, a dozen of them becoming, for example, Kazaa supernodes
would sap too much bandwidth from the network to allow us to
maintain good performance for the legitimate customers on that
portion of the network.

>  (You've stated elsewhere
>        that you run a wireless ISP.)  Alternatively, is the problem
>        that your network is underprovisioned for the load your
>        customers?

Our network is technically quite sound, thank you. But we need it
to be sound financially as well.

Remember, most ISPs oversubscribe their networks because they
have to. If they don't, their wholesale costs surpass what they
can charge at retail. We are one of the few that offer guaranteed
minimum throughput, and to keep that promise we cannot oversubscribe
the bandwidth which is "spoken for" when we make this guarantee. It
must be available to the user at all times. However, users will be
dissatisfied if they do not get substantially more than the
guaranteed minimum most of the time. This, again, becomes a duty
cycle issue. If users' computers do not attempt to monopolize
bandwidth, everyone is happy and we can break even and perhaps even
turn a small profit for our hard work. But if the pool of burstable
bandwidth is monopolized, our network will be perceived as slow,
users' productivity will be hampered, and they will leave.

File piracy software tries, by design, to take ALL of the available
bandwidth if it can. It tries to consume not only all of our burstable
bandwidth but attempting the guaranteed bandwidth that is reserved for
other subscribers. This constitutes network abuse.

>      In that case, is the problem economic -- the
>        market won't let you charge enough to cover your costs?
>        If there's a market problem, could there or should there
>        be some intervention to correct it?

Perhaps a market intervention called "truth in advertising" would help
a bit. When the phone company claims that you'll get 7 Mbps downloads
on your DSL line because that's the raw speed at which the modem
trains up (and then disclaims that you will get any speed at all in
the fine print), it is deceptive.

If users want a dedicated T1 to the Internet, they should be prepared
to pay the $500 per month that it costs wholesale in our area.

However, oversubscription of bandwidth works just fine, and allows users
to get a better deal, if they are not abusing the network.

>      Or is there some
>        technical shortcoming in your ability to do traffic-shaping

No; we can and do shape traffic. In fact, we have spent hundreds of
hours developing techniques for mitigation of abuse by file piracy
software.

>        or usage-sensitive pricing, in which case innovation may solve
>        your problem?

The fact that the market will not tolerate usage-sensitive pricing is
not a technical issue.

>        Should ISPs (including the university in its role as an
>        ISP for faculty, staff, and students) have the responsbility
>        for proactively blocking illegal traffic?  If so, how can
>        they distinguish between, say, a stolen Time-Warner movie
>        and the legitimate copy that's being redistributed via
>        BitTorrent at Time-Warner's explicit request?

Even the occasional legal use of BitTorrent can harm the network,
because (again) it attempts to monopolize it and thus is abusive.

>        Do we give up one of the fundamental tenets of the Internet
>        architeture, the notion that endpoints determine what
>        traffic flows, rather than the center?

In real life, the endpoints of the network have never completely
determined what traffic flows or how much.

>  In this regard,
>        it is worth remembering that the three most radical Internet
>        innovations -- the Web, Napster, and Skype -- came not from
>        ISPs, "official" standards bodies (i.e., the IETF or the
>        ITU), or major research labs or universities, but from the
>        edges of the Net.  (Yes, I know that CERN is a major research
>        lab, but for physics!)

The Web was not a radical innovation but rather an evolutionary step.
Napster and Skype likewise weren't radical. I daresay that wireless
broadband as we do it is more radical and revolutionary than any of
these. It was not supposed to be possible; you weren't supposed to be
able to break the telephone or cable monopolies or to offer a
commercial service without buying prohibitively expensive licensed
spectrum (which the big guys could lock up). But we're doing it.
All we ask is to be able to eke out a living -- something which we
at least hope that those enthusiastic customers want us to do. After
all, they do value our service.

--Brett Glass



-------------------------------------------
Archives: http://v2.listbox.com/member/archive/247/=now
RSS Feed: http://v2.listbox.com/member/archive/rss/247/
Powered by Listbox: http://www.listbox.com