May 22, 2008, 11:03 PM —
I have been thinking a lot about autonomous, software agents recently. I was somewhat startled when I concluded that we might be well on the way to creating a large number of these things, using nothing more complicated than Web technologies. Here is my (possibly) erroneous reasoning, expressed in flashback form. Some liberties have been taken with the timing of events, hopefully without detracting from the overall point.
In the beginning, there was the Web and it was good. However, it pretty much would just sit there in the ether until a human or a process (collectively a "user agent") would come along and ask to "see" something, generally in HTML or PDF form.
Time passed. It soon transpired that responding to requests from user-agents by generating something on-the-fly added a significant amount of power to the Web. Soon, the whole Web became engulfed in application frameworks for generating content on demand. It too was good.
Time passed. It transpired that a significant number of user-agents were not in fact, people and therefore it was much more useful to send machine readable information over the wire. XML, JSON, RDF, YAML etc. were all created to meet this need. They too were good.
Time passed. It transpired that a serious amount of effort was going into automating the process of change detection on the web. Feed formats and protocols came into existence that allowed applications to sit back and be notified when something of interest occurred. Formats like RSS/Atom were joined by protocols like XMPP and LLUP. They too were good.
Time passed. The change notification formats/protocols became increasingly sophisticated and machine-oriented. Soon, the major use cases for XML, JSON, RDF, YAML formats were event-oriented. At first, the primary consumers of these feeds were end-user applications which, in effect, mirrored the client-server model familiar to decades of users and developers.
However, still more time passed. XML 2.0 and HTML 6.0 walked the earth. More and more event-oriented applications came into existence that did not directly interact with end-user at all. Many of the client programs consuming event feeds became server programs, living in the cloud - just like the event servers they were interacting with. Statistics for web servers began to show dramatic growth in server-to-server communication with client-server communication falling significantly in percentage terms.
Nobody could pin-point when the so called "agents" running out there in the cloud became critical to the digital health of the planet. We were mostly taken by surprise. Our dependence crept up on us. Nobody had anticipated that some simple messaging protocols and data formats on the Web would have such a profound effect on the world.
Here we are in the year 201x. Time still passes and still we wonder as we watch the time whiz by. Did we create this increasingly intelligent, self-healing, evolving Web? Or did it cause itself to be created by the original, speed-challenged, scalability-impaired carbon-based agent technology know as Man?
