September 6, 1997

Letters

Give Dad his due

The mitochondrial DNA study of Neandertals rules out only maternal inheritance of this DNA by modern humans ("Neandertals make big splash in gene pool," SN: 7/19/97, p. 37). Until male-linked DNA analysis is possible, some degree of Neandertal ancestry for modern humans through the male lineage cannot be ruled out.

Lynn Hawley Bootes
Redlands, Calif.

Congestion suggestions stir storms

Concerning "Internet congestion stirs up data storms" (SN: 7/26/97, p. 53) and its proposed solution to usage spikes, the assertion that equations exist to describe such phenomena does nothing to validate either the equations themselves or the assertion that data charges are the way to go. The fundamental assumption that everyone, or even a significant proportion of users, will suck up bandwidth all day because it's free needs some closer examination as well. Let me mention a few actual problems.

First, it is not the amount of usage that creates congestion spikes, it is the fact that the usage is random.

Next, users indeed do not differentiate between low and high data transmission requirements. For one thing, the advertisers who insist on swamping us with elaborate graphics at every turn of the page are not charged in proportion to their excesses. How will a pricing algorithm distinguish between wanted and unwanted data? For another thing, it is a mistake to suppose that the average user will base his or her demand for service on charges for use of the available bandwidth. The average adult will use as much as seems necessary -- or in the case of teenagers, what they can get away with.

Finally, the article cites spikes that are frequent and last a fraction of a second. They must be really, really frequent for the user to notice them without specialized electronic equipment. My impression is that the server addressed gets swamped long before the transmission lines do.

As for the answer, it is provided in the article. Users tend to avoid delays by trying later or by timing their access to periods of low usage. This is called smoothing, and it happens automatically.

We already have a somewhat regressive pricing structure for the Internet, and we don't need one that favors institutional users (who are unlikely to lose their free ride in any case). Can you imagine a grad student paying punitive charges for graphics essential to his scientific endeavors? Or the government saving any money at all?

Last, how will an equitable pricing structure be implemented? First, you have to get all the current service providers to agree on a new -- and marketable -- scheme that does not violate restraint-of-trade laws. What will be done about overseas providers?

Any change in pricing will most likely evolve as a result of environmental pressures, not as a result of anyone's concept of what would be nice.

John Hannah
Silver Spring, Md.

The issues involved in flat-rate pricing are certainly not new (consider buffet lunches, unmetered phone calls, freeway use), and the occurrence of congestion should surprise no one.

The future of routing is probably a pay-for-priority scheme. Though billing per byte sounds satisfyingly simple, the nature of the World Wide Web in particular makes this problematic. It is easy to estimate costs for sending N bytes of E-mail, but the cost of connecting to many of today's graphics-heavy sites is impossible for the end user to estimate.

While there may eventually need to be a surcharge for multimedia transfers, a show of restraint on the part of Web designers should reduce much of the day-to-day congestion. Perhaps a picture is worth only 10 words on the Internet.

John M. Vinopal
Berkeley, Calif.

It's easy for the researchers to suggest that users are to blame for Internet congestion and should be charged according to the data they use. The researchers' own usage is paid for by the grants given them by Xerox and the University of Michigan.

I can understand Xerox's interest in the study. As for public university studies which suggest the public should pay more and more fees for using a public resource, maybe taxpaying users should conduct a study that investigates such research studies.

Maybe a scalable grant scheme is in order. I think a wise and sound approach would be to measure those studies that suggest "improvements" which increase (worsen) users' quality of life while saving (costing) money and then grant the studies more or less research money accordingly.

As it stands now, because researchers do not personally pay costs linked to their performance, they seem unable or unwilling to differentiate between activities that may require users to spend more money and irresponsibly thinking that their actions have little effect on the users.

Nancy Jane Mathews
Davis, Calif.

I have had many years' experience managing large segments of the Internet. I agree that the short-lived storms observed are normal, but I take issue with the conclusion that somehow charging for access (or altering the current charging scheme in some undefined way) will make the storms go away. The article presents no evidence linking the conclusions to the data.

Craig A. Finseth
Saint Paul, Minn.

Send communications to:
Editor, Science News
1719 N Street, N.W.
Washington, D.C. 20036
or: scinews@scisvc.org
All letters subject to editing.