Interesting People mailing list archives
aint we been here before
From: David Farber <farber () central cis upenn edu>
Date: Thu, 16 Sep 1993 11:17:10 -0500
Panel Gets task of closing gap between major comm protocols {Government Computer News}, September 13, 1993, Page 3 By S. A. Marco, GCN Staff The National Institute of Standards and Technology this week is expected to appoint a nine-member committee that will try to bridge the abyss between the TCP/IP and GOSIP worlds. The group has until Dec 31 to come up with answeres that so far have eluded the federal government and the network industry. Although the Government Open Systems Interconnection Profile was mandated three years ago for federal data communications procurements, agencies remain frustrated by the dearth and the high prices of GOSIP products. In contrast, TCP/IP products have grown more and more popular. But TCP/IP developers, whose most notable achievement is the worldwide Internet, are suffering from their own success. By the year 2000, the Internet probably will not be able to handle any more users. It's running out of IP addresses. Ironically, the GOSIP model's Connectionless Network Protocol (CLNP) could offer virtually unlimited addressing capability. It is at the network layer, where both IP (the Internet Protocol) and CLNP operate, that some people believe both camps can find a common ground. OSI TCP/IP +--------------+ +--------------+ | Application | | | +--------------+ | Application | A potential convergence | Presentation | | Presentation | | Session | | Session | between the OSI and TCP/IP +==============+ +==============+ [ Transport ] [ TCP ] stacks could result in ===============+ +==============+ [ Network ] [ IP ] a single set of protocols +==============+ +==============+ | Data Link | | Data Link | for the network and +--------------+ +--------------+ | Physical | | Physical | Transport layers. +--------------+ +--------------+ "The crucial thing, the real nugget in this, is not the fact that NIST is thinking of modifying GOSIP to accept TCP/IP," said Stephen Wolff, director of the National Science Foundation's Network Division. "The important part is that NIST is thinking of modifying GOSIP to provide for the convergence of the two protocol stacks. If all you do is accept TCP/IP as an alternate stack, then you end up with two non-communicating protocols," he said. Wolff is a member of the Federal Networking Council, which together with the Federal IRM Policy Council has submitted a lit of candidates for the nine-member Panel on Government Networking Policies. On the government side, some of the impetus for stack convergence is coming from the Energy Department, which works closely with industry, academia and researchers. Deceptive tranquility "We run both sets of protocols on our [Energy Sciences] network," said Robert Aiken, who is in charge of DOE's network research program. "It costs a bit of money to support two stacks." Aiken said users are more concerned about things like exchanging electronic mail and "99.99 percent don't care what the transport mechanisms are. But we seem to be hung up in the wars underneath, at the transport and network layers." Wolff believes the panel should focus on the network layer. "Then you can worry about trying to converge the transport layer, if you want to," he said. "If you converge at the network layer, then you can route around the world on a single network." Richard des Jardins of the GOSIP Institute in Fairfax, VA., said the government bears a heavy responsibility for the TCP/IP vs. OSI conflict. "Aspect No. 1 of this whole problem is that the government spent money to develop both of these suites," des Jardins said. "Now they are both fighting each other, so the government should spend the money to put them together and have only one suite. People don't want to have to buy a whole new network just because applications are from another suite." In a white paper issued in June, des Jardins detailed the GOSIP Institute's plan for a worldwide Internet based on protocols borrowed from both stacks. As a near-term solution, he proposes running TCP and its sister User Datagram Prtocol over OSI's CLNP - also referred to as "TCP and UDP with bigger addresses" or TUBA - and adopting a routing method called Classless Interdomain Routing (CIDR). "TUBA provides a way for OSI interests to support TCP/IP applications, specifically Telent and FTP" [File Transfer Protocol], des Jardins said in the white paper. "CIDR provides a way for IP addresses to last a very long time, at least through the end of the century." These ideas and a handful of others currently are being debated by the Internet Engineering Task Force. But the GOSIP Institute paper suggests that for the longer term, both the Internet and OSI communitiies should develop a single internetwork protocol, which could take on the dual roles of being the next-generation IP as well as CLNP version 2.0. It's possible that the panel could concentrate its focus on the network and transport layers. Equally possible, it could go further up both the TCP/IP and OSI stacks and incorporate application-layer TCP/IP protocols within GOSIP's fold. Industry stress Officials at J.G. Van Dyke & Associates, an Annapolis Junction, Md., company that helps agencies make GOSIP transition plans, said the network industry is burdened by trying to support two stacks at once. "Industry is not opposed to OSI. Industry supports TCP/IP because the development costs are lower," said Laura Boyer, a senior systems engineer for the company. "If the two suites were harmonized, then vendors wouldn't have to make that choice. They wouldn't need two product lines." To some TCP/IP protagonists, the panel's mission already seems doomed. "People who are stuck with GOSIP have my sympathy, but they can always jump ship," said TCP/IP guru Marshall T. Rose, a principal of Dover Beach Consulting, Inc., Mountain View, Calif. Rose noted that although the government has tried to stimulate the market for OSI products through its standards process, leading TCP/IP vendors such as FTP Software Inc. of Cambridge, Mass., have not gone over to building OSI products. FTP Software instead has responded by improving its TCP/IP products and cutting prices, he said. "Pragmatism tells us that these technologies, though superficially related, have radically different philosophies about networking and radically different ways of being standardized and produced," Rose said. "Thinking there's some way you can harmonize them in folly," he said.
Current thread:
- aint we been here before David Farber (Sep 16)
- <Possible follow-ups>
- Re: aint we been here before David Farber (Sep 16)
- Re: aint we been here before David Farber (Sep 16)