kwe@bu-cs.BU.EDU (kwe@bu-it.bu.edu (Kent W. England)) (11/19/88)
The subject is funding national and regional education and research networks and is prompted by Barry's challenge to post something worthwhile on the subject. Funding the Internet is a topic of great interest to me and has been discussed in the past on tcp-ip and the bitnet big-lan list in particular. These ideas aren't necessarily new, but I hope that a little data from the trenches will be useful. There are many committees and organizations echoing Gordon Bell's call in IEEE Spectrum to build a national, government sponsored, education and research network. Apparently, Albert Gore and Rep. Waldren (I forget his first name) are behind the idea in Congress and are simply asking for reasons and costs. There is a group of government agencies (NSF, DOE, DARPA, NIH, NASA, ...) that are working together to integrate their currently independent networks. There are at least three main thrusts behind these and other initiatives: coordinate networks and save on overall cost become more competitive in research and trade (compete with the Asians) spread the benefits of networking to a wider audience. While it's interesting to talk about the propriety of government involvement in subsidizing networking, it is a fact that they are already in the business of doing so and are interested in saving some of their current expenses. Witness DARPA trying to save money migrating ARPAnet customers to NSFnet regionals and the origin of the DRI, Defense Research Internet. This is all part of a trend to make networks more production oriented and save the research money for research. The other two thrusts are a result of the same thing; make these operational networks more production oriented and less research-oriented. Take the technology and use it in support of scholarly scientific and engineering research, not just computer science research. The drive is to get to production networks from where we are today to where they should be. Most of the government agencies want to get out of the networking business altogether and turn networking over to "someone else". Question is; How to do this? A direction, if there is one, is that NSF, DARPA and other agencies can transition the Internet to independence or quasi-independence over a relatively short number of years. This means that those institutions eligible to join the Internet (as defined by NSF, DARPA, et al) will have to begin to pay directly for networking service. There may be short term subsidies, direct or indirect, but the clear trend is to put the cost of networking back in the budgets of the schools and the research labs using the network. Institutions need to begin to think what networking is worth to them. But, you say, we aren't yet networked, or we haven't yet built the infrastructure on campus to exploit these national and regional networks. Good point. That's why some people want to subsidize networking a while longer until a "critical mass" is built and the Internet (used loosely to mean the NSFnet and regionals and what they evolve into) can move to cost-effective independence. What do these cost figures look like? Well, on the low end BITnet and CSnet sites pay $2k or around $5k plus line charges and on the high end, it's subsidized for the most part and we have no true cost/benefit analysis. But this is rapidly changing. High end users face unsubsidized or partially subsidized charges of from $30k-50k (and more) in the near term. Is networking worth it? What do we get for the money? e-mail, anonymous ftp, login, ftp in that order (IMHO). I can tell you if we move too quickly to unsubsidized networking we will freeze the mean at CSnet technology and not NSFnet technology becasue we don't yet have the critical mass. If we move more slowly, as there is every indication we will, then perhaps we can build that base that will support the relatively high cost of high speed connectivity which will allow more interesting network services to grow. I believe in government subsidies to grow networking to a critical mass. I think the national highway system is a good model, but I hope the "states" do a better job of maintaining the network resource than they did with the highway resource. I also believe that we should have a national production network for the exchange of scholarly and research information that will be self-sufficient and perceived as cost effective by those who use it and pay for it. I also believe we must continue research on networking and that research networks do not fit the production model above by definition and so should not be self-supporting (a term that is essentially meaningless when applied to research networks). How will these self-sufficient networks recover their costs? I leave that to another discussion. It's essentially a question of picking an appropriate cost model and an appropriate service policy. Kent England, Boston University