Site Home : Research : One Article
When you are talking to a tenured CS professor, you are talking to someone who has achieved his or her success under the 19th-century system of journal publication. Vannevar Bush noted in June 1945 that "publication has been extended far beyond our present ability to make real use of the record" ("As We May Think", reprinted in Computer Supported Cooperative Work (Greif; Morgan Kaufmann)) and Bush's proposed solution, the "memex", is very similar to the Web circa 1998. Nonetheless, if what you want is a lifetime guaranteed salary and students paying $26,000/year to hear you talk, it doesn't pay to subscribe to Bush's proposed innovations or Berners-Lee's actual innovation.
The primary goal of a would-be successful academic is journal publication. This is what leads to tenure. It does not matter if the academic (or anyone else) reads the journal in question. The tenure committee will note the successful navigation of the peer review process.
Because the goal of a computer science professor is journal publication, the most exciting computer applications for a computer science professor are Adobe Frame, Emacs, Latex, and Microsoft Word. People are not going to get excited about a computer program for which they have no use.
Fundamentally, the average computer science professor is not going to be interested in the Internet or database management systems. Suppose that you're a business executive and recently spent $200 million on an information system. It will irk you that you pay people to sit in front of terminals, talking on the telephone with employees of your customers and vendors who are sitting in front of their terminals. You will say "this $200 million computer system sucks; people are rekeying data all the time" and be receptive to the idea of an Internet in which structured data are exchanged between loosely cooperating enterprises.
A professor, however, is not a business executive. A computer science professor's computing needs are beautifully addressed by Microsoft Office.
Does this mean you're out of luck and will have to work on whatever projects your professors thought were interesting five years ago when they applied for grants? No. The federal government and corporations are heavy users of computing, which is a multi-$billion industry. All users of computing technology are passionately interested in Internet applications, which have the potential to finally deliver on the promises made for this technology since the 1940s.
You will have to demonstrate to your professors that your research has cash value for them, i.e., that they can somehow make money from it. This could be a research grant from the government or a corporation. This could be via commercialization of whatever you develop (though personally I think your own career and the world will generally be better served if you give your code away).
In the 1960s, top computer science research institutes such as Stanford and MIT had computing facilities comparable to those in large corporations. Researchers often did practical experiments. Today, however, you will probably find that a computer science department does not actually have any computers. They are amply supplied with word processors but don't have anything that you might realistically want to use for running a modest db-backed Internet service.
Sidenote: when my personal Web server, a mid-range HP Unix server
(K460), was rolled into the MIT computer science lab, professors were
aghast that anyone would need that much computing power. What could I
possibly do with 4 GB of RAM and 36 hard disk drives? They grilled the
HP field engineers who set up the machine:
Professor: "How many HP machines of this size are there in the world?"
HP engineer: "I have no idea."
Professor: "How many are there in the Boston area?"
HP engineer: "I really couldn't say."
Professor: "How many have you ever personaly seen?"
HP engineer: "Well, I was over at State Street Bank today and they have
50 machines just like this in one room. Gillette has 35 to run their
SAP system."
More difficult than getting computer hardware is the acquisition of relational database management system (RDBMS) licenses and the full-time personnel usually required to run them. Computer science research departments usually work with the data storage technology of the 1960s: the file system. File systems are great if your data is of no great value and you only have one user at a time making changes. However, the latter condition, at least, will be impossible to satisfy for any interesting Web service (see the db chapters of my book).
You can try to convince your fellow researchers of the virtues of the DBMS advances made in the 1970s. However, this will probably be difficult, especially as most of them either won't know what Oracle is or won't have ever typed a line of SQL in their lives. People who have no idea how to use Oracle aren't going to want to spend money on a pile of hard disks, an Oracle license, and dba expertise.
Building up your own computing facility from scratch is going to take a lot of time. Since your professors were happy with Microsoft Office, they aren't going to give you academic credit for setting up an Oracle server. Therefore, as far as your degree program is concerned, all of this is wasted time.
An alternative is to work with a commercial partner. The modern infrastructure that you need to build a pathbreaking Web application is ubiquitous inside Corporate America. Morever, IT managers and Web publishers at companies will be much more current with technology and the Internet than folks in academia. You will probably be able to explain the merits of your ideas in just a few minutes. The marginal cost of operating your service on their infrastructure will be very low. They may be interested enough in your results that they'll even pay you to use their computers!
Another kind of partnership to explore is working with folks at other universities, something that is more practical these days thanks to the Internet. By virtue of working with professors at University of Washington, for example, my friend Ellen had her choice of two maintained RDBMS installations when doing her (very interesting) Ph.D. research.
Because the Internet and the Web represent great opportunities. When I started my professional programming career in 1978, it was essentially impossible for an individual to make a contribution to society without working in a large organization. To contribute, you were forced to work on a project with 30 other engineers, typically for at least 3 years, in hopes that a salable product would result. Even if you achieved all of your technical goals, the company often decided that the costs of marketing the product would exceed the revenue. Your work never saw the light of day.
With the Internet, if you can get your hands on a properly maintained Oracle server, you can build a powerful Web application in weeks or months. By yourself. Your work can be experienced by 10,000 people within a few days of completion. People who never have to go through the pain of downloading and installing software.
Since nearly all of the folks in our CS department and Media Lab were openly scornful about the revolutionary possibilities of Web applications, I teamed up with Hearst Corporation in 1995. That gave me access to their modern computing facility, which I used to test out early versions of my community software ideas.
I couldn't resist the temptation to build my own computing facility here
at MIT and that probably cost me a year. However, I think it was worth
it because I have ample computing power and in-house expertise for my latest and greatest scheme.