August 17th, 2006

classic cylon

(no subject)

The thread about well-known ports on the IRTF end2end list has burst into a flame war about the lack of a complete Internet addressing architecture. Apparently, ARPA (the funding agency) did not want to fund continued work in that area. There has been a complaint that computer science researchers have not risen to the challenge of completing the architecture. In the course of the thread, I mused that a possibility for this is that most CS students seem to be gravitating towards other subdisciplines of CS such as data mining and information retrieval because they are used by search engines which are popular places to work. My former UROP advisor and 6.033 instructor gave an impassioned reply. He blames academia for trying to turn computer networking (and CS in general) into a discipline. (I may comment on this in this journal later.)

Actually I have always thought computer networking didn't have much student appeal, even as an undergrad. A lot of CS students gravitated towards AI projects. It is possible that AI projects were better promoted via 6.001 and 6.034, which some people could complete by the middle of sophomore year. OTOH one wasn't likely to encounter anything computer network-related until 6.033, something usually taken by juniors or seniors. People are discouraged from taking it earlier because the course doesn't strictly build on previous subjects; it requires a lot of "street knowledge". The same is true of 6.033 today, although it's possible to finish 6.001 and 6.034 by the end of freshman year. But then, as now, it seems a lot of students find topics like graphics, multimedia, AI, etc. more fun. (BTW, if you want to find out more about these classes, visit the EECS OCW page.)

Another reason other subjects are more popular could be that you get to interact with the software directly. Generally speaking, the networking code is either part of the OS or a special library. There are some UIs that allow you to get visual graphs of network devices and traffic, but again, one doesn't usually encounter that sort of program early in one's undergrad studies, unless one takes a special interest in networks. (There are some network simulation programs, but these have mostly to do with the analytical side of networking.)

These issues aside, I'm not sure why computer networking would be intrinsically uninteresting to students. But if the type of creative energy that goes into creating companies such as YouTube, MySpace, Google, etc. is drawing the crop of students who might otherwise go to Lucent or Cisco, we may not see the next generation Internet addressing architecture for quite some time.