Simple Wins

The message to learn from Ross Perot's career is think simple.

Before Perot ran for president, he made a lot of money offering computer services to companies. These companies either didn't want to make the investment in in-house computer processing, or did not trust their own DP department to perform mission-critical tasks. So companies paid Perot an awful lot of money to process their payrolls and claim forms, using simple, standard technologies.

---

OUTSOURCING THE INTERNET

Switch to today. Companies are starting to realize increasingly that mission-critical tasks involve interactions with other companies on the Internet. That is, an increasing number of business transactions are Internet-mediated. Although, these transactions may occur over Virtual Private Networks or extranets, the technologies are essentially the same.

Many of these companies lack faith in their MIS departments to deliver the required technologies, or they have come down on the "buy" side of the make/buy issue for straightforward economic reasons. Either way, they would like to outsource the processing of these Internet-mediated transactions (IMTs).

Here's an example close to home. The IEEE Computer Society is, of course, a publisher. Like many publishers, we use Quark to do electronic layout. While the Computer Society FTPs our Quark files to the printer, many publishers do not. Enter WAM!NET. They provide a turnkey service for file transport - publishers and printers don't even buy software. The monthly service fee covers everything. The publisher clicks on an icon when the Quark file is ready, the file is transferred between machines, and the printer is notified.

I can hear you saying, "But I could do that with a Perl script!" What do you think WAM!NET uses, albeit a complicated one? Most importantly, they provide a service. And they are an example of what is going to be one of the biggest type of businesses on the Internet: the outsourcing of IMT applications.

---

ENTER XML

These IMT apps are likely to explode with the deployment of XML for e-commerce. Yes, I know that my last column, "The XML Files," disparaged XML, but XML has momentum, and startups and consortia are already developing applications that define de facto standards and business ontologies. Concepts such as "purchase order" will become standardized and formalized using XML tags.

This standardization will facilitate businesses exchange documents that are both interoperable and human and machine readable. There will be a growing market for IMT services that provide the processing of these new e-commerce applications.

---

WHY AGENTS WILL LOSE

Some people have made grand predictions (for example, Peter Fingar's article, "A CEO's Guide to eCommerce Using Intergalactic Object-Oriented Intelligent Agents") about how agents will enable e-commerce.

But take a look at all the existing e-commerce systems and try to identify "traditional" academic multi-agent technology. It's not there.

After years of research, this year's International Conference on MultiAgent Systems showed again that the multi-agent community has yet to agree completely on a standard for exchanging messages, and there are no examples of two independently developed systems working together. This is in addition to the fact that the mobile agent community has yet to make a convincing case for an application of that technology.

E-commerce is happening without such agents. There are already commodity auctions such as FairMarket, and brokers such as FastParts.

Business-to-business IMTs are on the rise. And the technologies required to develop these apps are much more likely to be Perl, XML, Java RMI, and CORBA than the more complicated and less finished ACL technologies of KQML or FIPA. Yes, there may be "agents" of the simple WebCrawler/Shopbot/Jango/BargainFinder sort, but these use the former rather than the later technologies, and are not multi-agents in the sense of persistent decision-making entities cooperating using any of the academic protocols.

This is not to say that agents could not be useful in e-commerce. Various sorts of agents are quite likely to be useful in areas such as telecommunications (see the Intelligent Agents Telecommunincations Applications 98 conference). KQML may be useful as may Stanford's own JATlite agent infrastructure.

Rather I am saying that agents (of the academic complex sort) are not in the technical path of e-commerce firms and startups. If the agent community wants its grand e-commerce visions to happen, they need to do something now - in the next 12 months - to inject their technology into the emerging IMT standards and systems.

Meanwhile, people will be making money in e-commerce with Perl scripts and XML parsers, offering services instead of visions.

Just imagine, our next, interesting third-party presidential candidate may have made their fortune as an XML geek.

---

URLs in this article

BargainFinder bf.cstar.ac.com/bf/

"CEO's Guide to eCommerce" home1.gte.net/pfingar/eba.htm

FairMarket www.fairmarket.com/

FastParts www.fastparts.com/

IATA 98 dai.cs.tu-berlin.de/workshops/IATA98/iata98.html

International Conference on Multi Agent Systems www-leibniz.imag.fr/MAGMA/ICMAS98/

Jango www.jango.com/xsh/

Jatlite java.stanford.edu/

Shopbot www.shopbot.com

"The XML Files" http://cdr.stanford.edu/~petrie/online/v2i3-webword.html

WAM!NET www.wamnet.com/

WebCrawler http://www.webcrawler.com/