Forum

In "It’s Time to Think Outside the Computational Box" (Nov. 2005), Peter Kugel suggested we could increase computing power by thinking beyond the narrow model of the classical (functional) computation of Turing machines. He pointed to the limitations of learning through predefined input and immediate output. He suggested "computing in the limit" (CIL) through unlimited input and delayed multiple output, modeling broader computational problem solving. He showed that CIL allows computers to express "programming by example" through unlimited observation to determine correct understanding (learning) of examples comparable to human understanding.

Though we agree that the Turing machine is a limited model of computation, many ways of thinking outside the box remain beyond Kugel’s extension of the limits of computing. CIL expands problem solving to include programming by example but does not adequately handle broader problem solving through learning and intelligent agents. For example, in reinforcement learning, an agent must actively explore its environment, not passively observe it.

Interaction is an extended form of computer problem solving that breaks out of the "computational box" by viewing computation as an ongoing process that consumes inputs and produces outputs by interacting with its environment to generate the desired computing behavior (such as reinforcement learning).

Interactive computation interleaves dynamically generated streams of inputs and outputs, thus supporting a stronger extension of problem solving than CIL. Both CIL and interactive computing are paradigms that change the underlying model of computation by thinking about problem solving very differently from Turing machine models and from each other. (Our forthcoming book Interactive Computing: The New Paradigm will focus on interactive computing outside the box of Turing machine computation.)

Alan Turing developed his Turing machine model (before the advent of computers) to show that the mathematical decision problem was unsolvable. His work was reinterpreted in the 1960s as a complete model of computational problem solving. This reinterpretation excluded thinking outside the box, because alternative problem-solving models could not be more powerful. Since then, new applications of computing have created a range of novel forms of computing behaviors that exhibit learning, concurrency, and other modes of problem solving.

Interactive computation may turn out to be more useful than CIL because it expresses a broader form of computing. Meanwhile, further analysis of both extensions to classical computation will contribute to a better understanding of the fundamental principles of computing.

Thomas Kuhn encouraged thinking outside the box in physics and mathematics as a form of scientific revolution where new paradigms expand understanding despite initial resistance by proponents of the old paradigms. Kuhn’s analysis applies equally to computer science, encouraging us to think beyond the Turing machine and explore novel paradigms of computer problem solving.

Peter Wegner
Dina Goldin
Providence, RI

Lack of Ada Reflects Software Immaturity

In "Detection and Prevention of Stack Buffer Overflows Attacks" (Nov. 2005) Benjamin A. Kuperman et al. wrote "One way to prevent programs from having such vulnerabilities is to write them using a language (such as Java or Pascal) that performs bound checking. However, these languages often lack the low-level data manipulation needed by some applications." In this they ignored Ada, an industrial-strength language that includes bound checking, as well as other features, for safer programming while allowing low-level manipulations (such as bit twiddling, representation clauses, and interrupt management).

Many of the problems addressed in the article follow from not using appropriate programming languages, an issue Ada solved more than 20 years ago. Computer safety should be a result of good engineering, and the first step in engineering involves selecting the most appropriate tools for the task at hand. The fact that Ada is often ignored, especially in areas like those covered in the article, is a sign of the software industry’s lack of maturity.

Jean-Pierre Rosen
Paris, France

Be Skeptical of Rhetorical Slight of Hand

I appreciate the frustration that motivated Amy S. Bruckman to write her "Viewpoint" ("Student Research and the Internet," Dec. 2005). As scientific knowledge and access to it increase exponentially, we see a no-less- powerful reaction against it. Understandably (perhaps), that reaction must be troubling to anyone wishing to hold onto a long- held, deeply cherished belief. The column left me with the old saw that the truth is simple. The more correct an answer the more likely one’s response will be "Duh." Conversely, the more convoluted and prepositional the explanation, the more elaborate the justification and the more likely one has spun an invalid premise.

Even if the truth is simple, stating a blatant logical fallacy can still seem more satisfying. The most important skill students need to process the onslaught of information is not so much how to evaluate credentials and peer ranking but how to detect the rhetorical slight of hand of the skillful misuser of logic.

Curtis Rhodes
Farmers Branch, TX

More to Innovation than Innovation Alone

The kind of innovation Peter J. Denning and Andrew McGettrick proposed in their "The Profession of IT" column ("Recentering Computer Science," Nov. 2005) is not the only way to reorient computer science to make it more appealing. A better approach might be to relate CS to the panoply of non-CS human interests. Themes like "CS for visualizing global human trends and needs," "CS in business finance," "CS in archeology," or almost any other interdisciplinary marriage with CS comes to mind. We tend to learn best when motivated by personal interest. A student interested in the general concept of innovation would find the theme of innovation appealing, as the authors suggest. But many more of us have other interests and would be attracted to themes other than "innovation." A more incisive effort to recenter CS requires building relationships with those inherently diverse interests, rather than with just a single concern.

C.J. Fearnley
Upper Darby, PA

Authors Respond:

We agree that innovation is not the only alternative to programming as an organizing theme for CS degree programs. Others include CS-to-other relationships (the one Fearnley endorses), experimental CS, games, knowledge management, and the human-computer interface. In "A New Framework for Computer Science and Engineering," Paul Rosenbloom (in IEEE Computer, Nov. 2004) explored how this was being done at the University of Southern California where he is a professor of computer science.

ACM’s Education Board ought to identify which of them may have a following and develop advice for those wishing to try them out.

Peter J. Denning
Monterey, CA

Andrew McGettrick
Glasgow, Scotland

Handles Not a Naming Solution

I was amazed that Michael J. O’Donnell ("Separate Handles from Names on the Internet," Dec. 2005) didn’t consider the downside of his proposal, even as it didn’t solve any of the issues regarding domain name limitations. That proposal is an adjunct to the existing domain name schema; note his reliance on handle.root.nicesponsor.org. Anyone using this naming convention would still be at a commercial disadvantage vis-à-vis competitors that choose not to use it.

Using O’Donnell’s example, Grampa’s Candy store (candy.example) would be ZV9K1344OP.handle.root.nicesponsor.org yet still compete with Pete’s Candy Store (petes_candy.example) and MyCandy.xxx. Which of them would have the most visibility—a key to retail success, often described as "location, location, location"? Grampa’s Candy store is thus at a competitive disadvantage.

O’Donnell’s solution ignored the possibility of corporations redirecting subdomains at will under the existing system; they don’t need "handles." Moreover, his example of a domain name dispute is inaccurate. If Grampa’s Candy Store did indeed apply for and receive candy.example, Grampa would not be forced to turn over that domain. Domain disputes occur when large corporations claim trademark infringement or the original domain holder registered it in bad faith. Note that some organizations, including Microsoft, have been involved in domain disputes over sound-alike domains (such as MichaelSoft.com).

Part of the problem in the domain name scramble is that companies don’t register their domains once but under multiple TLDs, possibly .com, .org, and .net, so one domain might have multiple handles while simultaneously shrinking the pool of available domain names.

O’Donnell’s solution would actually make it easier for criminal activity to occur. To use a real-world example, fake Rolex watches are sometimes hawked via spam and throwaway domains and Web sites. One need only set up a series of throwaway domains that redirect to their actual domains, then advertise one of them per each spam run, keeping the spammer in business that much longer. Another example involves online pharmacies selling "drugs" via "affiliate" programs.

O’Donnell’s solution neither solves the domain issue nor eliminates the potential for more abuse. I don’t disagree that there’s a problem, but do find that O’Donnell’s solution is not viable.

Mike Segel
Chicago, IL

Author Response:

My proposal was not intended as a way to "solve any of the issues regarding domain name limitation” but as "an adjunct to the existing domain name” system, abandoning the valuable semantic associations of domain names to gain reliable permanence, reduced administrative costs, and verifiable resolution by end users.

There is no "reliance on handle.root.nicesponsor.org.” The best way to experiment with a system of self-assigned public-key handles is to implement it through the current DNS.

No "commercial disadvantage” is incurred through the handle system, though there may be a competitive disadvantage in shunning DNS altogether. But a user may, for example, register a domain name and also a handle, resolving the name to the handle.

I did not ignore the fact that organizations can "redirect subdomains at will under the existing system.” Rather, I proposed an added value for self-assigned end-user-verifiable handles in exchange for the reduced value from meaninglessness. Nobody needs handles. We must instead ask what kind of value free or cheap promiscuously self-assigned handles would give the citizens of the Internet, particularly those who find DNS registration too costly or troublesome. The answer may have to be resolved in court.

In 2002, Milton Mueller [of Syracuse University] reported "more than 6,000 cases involving over 10,000 domain name registrations” under ICANN’s Uniform Dispute Resolution Policy. I don’t know how many cases went to court or how many fraudulent acts misdirected domain names. The legal costs of DNS are at least in the millions per year. Whatever the outcome of speculative disputes over Grampa’s fictional candy.example, disputes are costly enough that it’s rational to want to avoid exposure.

I welcome serious analysis of potential abuses of a promiscuous handle system. Note that the impact of such a system on the total risk profile of the Internet depends not only on the system itself but on the way related systems (such as DNS) react to its presence. Note, too, that the verifiability of handles based on public keys may offer new defenses against these risks.

Michael J. O’Donnell
Chicago