Re: What is "normal Internet service"?

[David Farber <dave () farber net>]

Begin forwarded message:

From: Brett Glass <brett () lariat net>
Date: June 15, 2009 10:27:57 PM EDT
To: "Steven M. Bellovin" <smb () cs columbia edu>, dave () farber net, "Ip  
ip" <ip () v2 listbox com>
Subject: Re: What is "normal Internet service"?

At 06:24 PM 6/15/2009, Steven M. Bellovin wrote:

> Brett, what do you see as "normal" Internet usage?

For what class of user? We have lots of different classes of service,  
each tailored to a particular class of user.

> You've spoken quite
> negatively (and quite often!) in this forum about peer-to-peer,

For good reason. Besides having been invented for the specific reason  
of facilitating piracy, it's a cost-shifting mechanism which shifts  
costs to ISPs without permission or compensation.

> bandwidth management,

An essential function for any ISP or enterprise. Bandwidth is a vital  
resource, and the most expensive one we buy. (I have just received the  
best offer I have ever gotten on backbone bandwidth: $76 per Mbps per  
month, and then only if we buy at least a DS-3 and lock the price in  
for two or three years. The really scary thing is that demand might  
force us to take it. We are seeing a bandwidth crunch in the evenings.)

> etc. Today, you spoke of "the bandwidth that
> YouTube can squander". But, like it or not, Youtube is one of the most
> popular applications on the web.

I'm not just being nit-picky when I note that YouTube isn't an  
application. It's a content provider. One of many with Web sites that  
host video. Adobe Flash Player is the application.

> The question, then, is this: what is
> the service that you and other similar ISPs intend to provide?

At a minimum, good connectivity. But that's a commodity, so to be  
competitive one must add value to it. I already add some value to it  
by offering delivery to remote locations, security, support,  
equipment, and consulting. But I don't want to be limited to that.  
Whatever I decide that I want to provide (after asking my users and  
weighing whether it's feasible), I'd like to be able to provide.

> What is your vision of what the Internet should be?

The Internet has always been what it is now: a federation of  
independent network owners who exchange traffic via standard protocols.

I do have a vision of how the Internet economy, the mix of  
applications delivered over it, etc. will evolve, but I'm not  
interested in forcing anyone else to conform to that vision (unlike  
some lobbyists who seem to want very much to do so).

> To me, you seem to be
> advocating the Internet of 10 years ago: relatively small transfers,
> comparatively static (and hence cacheable) web pages, a download- 
> mostly
> model, etc.

All of those things simply mean efficiency. (Are you suggesting that  
anything that makes efficient use of resources is obsolete?) In any  
event, efficiency is key to my network, because bandwidth is  
exceedingly expensive here. I can make an economic case for serving  
areas that the big guys cannot serve precisely because my network is  
efficient.

> I accept that your network is geared towards that, for both economic
> and technical reasons.  But it's quite clear that that's not the
> product most people want to consume (albeit perhaps not pay for).

Not true. Most people want basic Internet functions such as Web and e- 
mail. Some of them are leaning toward streaming video, but most are  
using the delivery mechanisms which are most efficient for video  
(satellite, CATV where it is available, over-the-air TV from our local  
analog translators) rather than trying to use the Internet for it  
(which is horribly inefficient at it; the cost is thousands of times  
greater per minute of video delivered).

> How can these be reconciled?

There isn't a conflict. If people really want to do something  
inefficient with the Internet, they can. But someone has to cover the  
cost, because bandwidth is quite expensive and we can't afford to  
waste it. Our basic residential services cost what they do because we  
expect to realize certain efficiencies. These include caching (to  
reduce our backbone bandwidth load) and moderate duty cycles (to allow  
oversale). Demand which violates these assumptions has to be handled  
via a more expensive tier of service, simply because it costs us more  
to deliver that service. We are not out to do anything evil; we just  
need to base prices on our costs. That's only fair.

> I see two alternatives here.  First, there is a regulatory approach:
> applications that don't meet a certain model of the network are
> banned.

We don't ban applications. However, we do have certain classes of  
service in which certain costly behaviors, at the network level, are  
prohibited by contract. For example, our residential classes of  
service prohibit the operation of servers, including things like P2P  
and the Slingbox. Our business class services make fewer (or no)  
assumptions about economy, and therefore are priced higher but do not  
have these restrictions.

The proponents of regulation seem to be bound and determined either to  
cause us to lose money or force us to raise our prices by forcing  
EVERY user into those higher tiers.

> This seems unlikely in most countries.  If nothing else, it
> would freeze innovation -- to me and to many others, the beauty of the
> Internet is that it encourages innovation at the edges.

To me, the beauty of the Internet is that it allows innovation  
EVERYWHERE. It is Ludditism to banish innovation from the middle of  
the network and restrict it to the edges. If I did not innovate in the  
middle, my customers wouldn't be able to get service at all at the  
edges; those edges wouldn't reach them.

> The other alternative is to let the market work its will.  If the  
> costs
> (and hence prices) for the service people want are too high, people
> will scale back their expectations or demands.  If others can offer a
> service you cannot or will not, then those others will prosper at your
> expense.

There's nothing wrong with that. Absent anticompetitive tactics or  
overly broad patents, we can copy one another's successful ideas.

> Your course, as best I can tell, has been load management: throttle
> expensive requests to something you can afford without raising prices
> too much.  Users may or may not be satisfied with that in the long
> term, but absent regulatory intervention the market will decide.

Users would be even less satisfied with higher prices.

> The one strategy you cannot pursue is to demand that others meet your
> definition of the Internet.

Oh? Google is pursuing that strategy, with altogether too much  
success, thanks to its deep pockets and "astroturf" lobbying in DC.

> Google presumably has its own reasons for
> not wanting Youtube videos cached.  Perhaps it's so they get an
> accurate hit-count, which they need for feedback and advertising  
> sales,
> and even for their "most viewed" counters.

That's absurd. Caches routinely poll for "freshness." That's the  
reason for "if-modified-since" queries in HTTP, which are just as easy  
to tally as full transfers.

> In other words, they're maximizing their own utility function.

Well, if they choose to benefit themselves by raising our costs a  
thousandfold, someone had better be willing to pay for the costs that  
result. Which will be substantial. A lot of our traffic at peak  
periods (and we must always buy for the peak) is YouTube. So, who will  
it be? The users or Google? It has to be one or the other. We're not a  
charity.

> It isn't reasonable to ask them to maximize yours.

I've tried to offer them a win/win proposition (caching saves them  
bandwidth, too), but they seem uninterested -- perhaps because their  
bandwidth costs are so much lower than mine. So, I have no other  
option but to offer them a choice: cover the costs or WE will limit  
those costs.

What's ironic is that Whitacre -- whom I at first thought was  
uninformed about the Internet -- is turning out to have been  
prophetic. Google is seeking to use our pipes --for all this  
unnecessary traffic -- for free.

What's worse, I predict that Google will, if it gets the ironclad  
"network neutrality" legislation that it wants passed, will start  
doing more and more of this. YouTube has never been profitable due to  
its high bandwidth costs. How much do you want to bet that Google is  
just spoiling to make YouTube P2P? (Note that they already got Adobe  
to build P2P into the latest version of the player.)

> You can respond -- but at a certain risk, as outlined.

We have to. We owe it to our users to keep our prices reasonable and  
to keep our business sustainable so that they are not left without  
service.

--Brett Glass





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