kwang@infmx.UUCP (Kwang Sung) (06/09/90)
[Part 1] Hi... Recently, I've proposed a new newsgroup "comp.protocols.migrate.to.iso". I've received so many responses. Among those, I've selected some best ones, after I've thrown out some bad ones. I couldn't reply to all of you, since it's so many. However, I might be able to answer thru this network. I think some people still don't understand how/where this networking world is going to. Absolutely, and positively, the world is moving toward one OSI world !!, even if ISODE has screwed up OSI world a little bit. Moreover, a lot of people were also interested in the network technology in Korea. I would like explain later as much as I know about it. I am sure those informations should be much more updated now than a year ago. Those informations were based on the systems/networks on KAIST, KIST/SERI, ETRI, RIST, Korea Defense Department, DACOM, SNU, and several industries in Korea. First of all, I would like to show some responses in order to prove why we need to create those new newsgroup. Here are some beautiful statements. I couldn't read them alone !! ------------------------------------------------------------------------------- >From: uunet!unx.dec.com!jjf (8N8-FRANEY) >Kwang and Marshall, >Our postnews is currently broken so sorry for the direct mailing. The topic you are discussing is so fresh and hot in my veins that I had to join in. I recently returned from a symposium in Washington DC that explored the transisition from TCP/IP to OSI. The word is transition since migration means travel to a certain place with intention to return. There will be no intention to return to TCP/IP once OSI is established. Firstly, the people who ran and participated in the symposium can easily be recognized as leaders in communication technology. Dr. John McQuillan - consultant and free lance author of about 200 articles on data communication, designed and implemented parts of ARPANET. Leonard Kleinrock - Professor of CS at UCLA, author of intense queing theory textbooks, member of Natinal Academy of Engineering, Guggenheim and IEEE fellow, etc... Dr Vintont Cerf - Vice President of the Corporation for National Research Initiatives, Chariman of Internet Activities Board, extensively involved with creation and growth of ARPANET. Dr. Jeffery Case - deisgnier of SNMP and other network protocols, professor of CS at U of Tenn. Larry Green - founder of Protocol Engines, Inc (developting XTP), working on networking technologies for the last 25 years. Bill Johnson - VP of Networks and Communications at DEC. Don Holtz - Products manager for IBM's OSI, MAP and TCP/IP communciations development. Bob Burnett - vp of Engineering for cisco. John Hart - VP of Engineering for Vitalink. etc... All that follows come from the discussions at this symposium. The major reason for a slower than expected transition to OSI is the installed base. TCP/IP is only a small part of the installed base with 1500 networks. PC Lans total over 1 million networks, IBM SNA networks control hundreds of thousands of terminals, DECnet has over 700,00 nodes (hosts and servers) and DEC has sold 3,000,000 ethernet ports to 61k customers. The argument that OSI is technically inferior to TCP is a myth. TCP and TP4 are practically the same (so much the same that people don't see a benifit of using TP4 over TCP.) Many of the TCP users will have the industry believe that the Dod protocols having been used and tested over the past 20-30 years is better than a set of protocols designed by open consensus. For an attack on TCP/IP, the experts (including Cerf) pointed out the TCP/IP was never meant to support a network as large as the internet. In fact, at the rate of growth of TCP/IP, it will run out of address space within the next 5-6 years. SNMP is an example of hind-sight engineering, as is sub-networks. These are features designed into TCP after years of operation. OSI has the advantage that it was able to use what was learned by TCP/IP people to design the entire protocol set. TCP/IP is an old technology. ALthough it has run on FDDI it wasn't designed for the high tech media. OSI also has limitations here but has afforded a modular stack to allow replacements of layers for progressive evolution. TCP/IP will not support the upcomming applications such as multimedia as well as OSI will. TCP/IP is connectionless, OSI is connection oriented. Multimedia needs multiple synchronized data circuits. Datagrams can't be lost or take variously different routes (TCP/IP). The major reason why TCP/IP is so prevalent in computer networks today is not because it is technically superior than anything that was put out before. But because it was given away as standard equipment on millions of workstations. It was part of the BSD UNIX that was sold on SUN boxes. If not for this fact alone, there probably would not be ANY country wide interoperable protocol. OSI is coming of its own. GOSIP is US govt's requirement list for computer procurements. Initially specifying a small subset of OSI, it will evolve to include the entire specifications. No company will win a GOSIP bid without OSI. And the US Govt is will start using GOSIP to describe contracts in August 1990. DEC has commited itself to OSI. All of its customers computer network nodes are going to be converted to OSI with the release of DECnet Phase V. I work for DEC but not in communications engineering but please don't think this is a plug. DEC is the only company to be committed to converting its installed base of proprieatary protocol to OSI. DECnet OSI gateways are expected to be released in December (don't quote me, I am not speaking for DEC here). With all DECnet as an OSI network, OSI will be on its way to being more accepted. Other companies are committed to OSI. IBM (good-luck), HP, UNISYS, cisco, Vitalink, RETIX (has implemented OSI in streams) and others. With corporate support, OSI will move faster. (A point made by McQuillan is that the Ethernet became a commercial success because research, standards and corporate communities came together. For OSI only standards community and more the corporate community have endorsed it. The research community is behind TCP/IP for good reasons). Corporations are developing transision strategies. There are many way to transision to OSI from any protocol. There are OSI gateways, application gateways, trasnport bridges and RFC 1006. Some customers are buying small OSI networks for experimentation only (this approach increases expertise before the network becomes critical.) TCP/IP is NOT popular in Europe. It doesn't surprise me that its not popular in Korea either. The big reason here is that the PTTs that implement and use the networks in Europe are connection oriented. They wand Virtual Circuits. TCP/IP is a datagram network and is not acceptable. Although there are TCP/IP node across the world, the North American Continent is where it is most poular. Europe, conerned about its 1992 common community has looked to OSI as common protocol since its inception. Europe is poised to take advantage of OSI and will have nothing to do with TCP/IP. Europe is composed of many PTTs that use X.25 and other CCITT protocols, not TCP/IP. The future of internetworking is not simple. The experts see TCP/IP as an integral part of the worlds networking infrastructure. But most will agree that the core of the infrastructure must be based on a globally accepted protocol, no matter how long it takes to implement. This is OSI. DEC sees OSI as an enterprise wide backbone, with gateways for proietary and early public networking protocols. Basically, the world is saying that any and all protocols are going to be uses. But the protocol itself is about important to most users as whether your car has efi or carbeurator that is not a significant amount. Users want interoperability. OSI, TCP, PC Lan, SNA, etc.. will be with us for a long time (10-15-20-30 years). The one that outlasts the rest will be the ones that best demonstrate interoperability. >Think long term. >Regards, >John Franey >These comments are generated from the ideas presented at the symposium >entitled: >TCP/IP & OSI, The Interoperability Challenge, Washington, D.C., >May 21-23 1990 hosted by the Technology Transfer Institute, Santa Monica, CA. >These ideas are not mine and I am not speaking for my company in any respect. -------------------------------------------------------------------------------- >From: att!lzaz!jer >I support your suggestion. >joe ritacco -------------------------------------------------------------------------------- >From: uunet!sunkist.West.Sun.COM!gene.saunders (Gene Saunders Sun SE Irvine CA) >I'd suggest the newsgroup be named "comp.protocols.tpicp.to.iso". >More descriptive. >Gene Saunders \ Gene.Saunders@West.Sun.COM \ 72265,23 (CompuServe) >Systems Engineer \ saunders@sunkist (local) \ GSAUNDERS (GEnie) >Sun Microsystems \ #saunder@ProteonW (from Novell) \ ..!sunkist!saunders -------------------------------------------------------------------------------- >From: Amit Parghi <uunet!watmath!watcgl.waterloo.edu!aparghi> >Organization: Computer Graphics Lab, University of Waterloo >Don't use the name "comp.protocols.migrate.to.iso"; since there are >already groups under "comp.protocols.iso", call it >"comp.protocols.iso.migration". >You claim the state-of-the-art engineers in Korea and Japan don't much >care for TCP/IP; instead, they use (I assume) ISO solutions, which I guess >would mean full ISO stacks implemented efficiently across heterogeneous >hardware and physical media. I, for one, would be very interested to hear >of such developments, especially since serious development of efficient and >*useful* implementations in Europe and North America still seems to be some >ways away. Do tell us more. -------------------------------------------------------------------------------- >From: Steven Willis <nsc!amdahl!ames!harvard!talcott!wellflt!swillis> >Consider this a vote for the newsgroup. There's an effort in the IETF and >ANSI to provide dual IS-IS routing for both IP and OSI. Ross Callon is >the chairman; you can reach him at callon@bigfut.enet.dec.com. -------------------------------------------------------------------------------- Hi... Recently, I've proposed a new newsgroup "comp.protocols.migrate.to.iso". I've received so many responses. Among those, I've selected some best ones, after I've thrown out some bad ones. I couldn't reply to all of you, since it's so many. However, I might be able to answer thru this network. I think some people still don't understand how/where this networking world is going to. Absolutely, and positively, the world is moving toward OSI world !! A lot of people were also interested in the network technology in Korea. I would like explain later as much as I know about it. I am sure those informations are now much more updated than a year ago. Those informations were based on the systems/networks on KAIST, ETRI, RIST, Korea Defense Department, DACOM, SNU, and several industries in Korea. First of all, I would like to show some responses in order to prove why we need to create those new newsgroup. Here are some beautiful statements. I couldn't read them alone !! ------------------------------------------------------------------------------- >From: uunet!unx.dec.com!jjf (8N8-FRANEY) >Kwang and Marshall, >Our postnews is currently broken so sorry for the direct mailing. The topic >you are discussing is so fresh and hot in my veins that I had to join in. >I recently returned from a symposium in Washington DC that explored the >transisition from TCP/IP to OSI. The word is transition since migration means >travel to a certain place with intention to return. There will be no intention >to return to TCP/IP once OSI is established. >Firstly, the people who ran and participated in the symposium can easily be >recognized as leaders in communication technology. >Dr. John McQuillan - consultant and free lance author of about 200 articles > on data communication, designed and implemented parts of ARPANET. >Leonard Kleinrock - Professor of CS at UCLA, author of intense queing theory > textbooks, member of Natinal Academy of Engineering, Guggenheim and > IEEE fellow, etc... >Dr Vintont Cerf - Vice President of the Corporation for National Research > Initiatives, Chariman of Internet Activities Board, extensively > involved with creation and growth of ARPANET. >Dr. Jeffery Case - deisgnier of SNMP and other network protocols, professor > of CS at U of Tenn. >Larry Green - founder of Protocol Engines, Inc (developting XTP), working on > networking technologies for the last 25 years. >Bill Johnson - VP of Networks and Communications at DEC. >Don Holtz - Products manager for IBM's OSI, MAP and TCP/IP communciations > development. >Bob Burnett - vp of Engineering for cisco. >John Hart - VP of Engineering for Vitalink. >etc... >All that follows come from the discussions at this symposium. >The major reason for a slower than expected transition to OSI is the installed >base. TCP/IP is only a small part of the installed base with 1500 networks. >PC Lans total over 1 million networks, IBM SNA networks control hundreds of >thousands of terminals, DECnet has over 700,00 nodes (hosts and servers) and >DEC has sold 3,000,000 ethernet ports to 61k customers. >The argument that OSI is technically inferior to TCP is a myth. TCP and TP4 >are practically the same (so much the same that people don't see a benifit >of using TP4 over TCP.) Many of the TCP users will have the industry believe >that the Dod protocols having been used and tested over the past 20-30 years >is better than a set of protocols designed by open consensus. >For an attack on TCP/IP, the experts (including Cerf) pointed out the TCP/IP >was never meant to support a network as large as the internet. In fact, >at the rate of growth of TCP/IP, it will run out of address space within the >next 5-6 years. SNMP is an example of hind-sight engineering, as is >sub-networks. These are features designed into TCP after years of operation. >OSI has the advantage that it was able to use what was learned by TCP/IP people >to design the entire protocol set. >TCP/IP is an old technology. ALthough it has run on FDDI it wasn't designed >for the high tech media. OSI also has limitations here but has afforded >a modular stack to allow replacements of layers for progressive evolution. >TCP/IP will not support the upcomming applications such as multimedia as >well as OSI will. TCP/IP is connectionless, OSI is connection oriented. >Multimedia needs multiple synchronized data circuits. Datagrams can't be lost >or take variously different routes (TCP/IP). >The major reason why TCP/IP is so prevalent in computer networks today is >not because it is technically superior than anything that was put out before. >But because it was given away as standard equipment on millions of >workstations. It was part of the BSD UNIX that was sold on SUN boxes. >If not for this fact alone, there probably would not be ANY country wide >interoperable protocol. >OSI is coming of its own. GOSIP is US govt's requirement list for computer >procurements. Initially specifying a small subset of OSI, it will evolve >to include the entire specifications. No company will win a GOSIP bid >without OSI. And the US Govt is will start using GOSIP to describe contracts >in August 1990. >DEC has commited itself to OSI. All of its customers computer network nodes >are going to be converted to OSI with the release of DECnet Phase V. I work >for DEC but not in communications engineering but please don't think this is a >plug. DEC is the only company to be committed to converting its installed >base of proprieatary protocol to OSI. DECnet OSI gateways are expected to be >released in December (don't quote me, I am not speaking for DEC here). >With all DECnet as an OSI network, OSI will be on its way to being more >accepted. >Other companies are committed to OSI. IBM (good-luck), HP, UNISYS, cisco, >Vitalink, RETIX (has implemented OSI in streams) and others. With corporate >support, OSI will move faster. (A point made by McQuillan is that the Ethernet >became a commercial success because research, standards and corporate >communities came together. For OSI only standards community and more >the corporate community have endorsed it. The research community is >behind TCP/IP for good reasons). >Corporations are developing transision strategies. >There are many way to transision to OSI from any protocol. There are OSI >gateways, application gateways, trasnport bridges and RFC 1006. Some customers >are buying small OSI networks for experimentation only (this approach increases >expertise before the network becomes critical.) >TCP/IP is NOT popular in Europe. It doesn't surprise me that its not popular >in Korea either. The big reason here is that the PTTs that implement and >use the networks in Europe are connection oriented. They wand Virtual >Circuits. TCP/IP is a datagram network and is not acceptable. >Although there are TCP/IP node across the world, the North >American Continent is where it is most poular. Europe, conerned about its 1992 >common community has looked to OSI as common protocol since its inception. >Europe is poised to take advantage of OSI and will have nothing to do with >TCP/IP. Europe is composed of many PTTs that use X.25 and other CCITT >protocols, not TCP/IP. >The future of internetworking is not simple. The experts see TCP/IP >as an integral part of the worlds networking infrastructure. But most will >agree that the core of the infrastructure must be based on a globally >accepted protocol, no matter how long it takes to implement. This is OSI. >DEC sees OSI as an enterprise wide backbone, with gateways for proietary >and early public networking protocols. >Basically, the world is saying that any and all protocols are going to be uses. >But the protocol itself is about important to most users as whether your car >has efi or carbeurator that is not a significant amount. Users want >interoperability. OSI, TCP, PC Lan, SNA, etc.. will be >with us for a long time (10-15-20-30 years). The one that outlasts >the rest will be the ones that best demonstrate interoperability. >Think long term. >Regards, >John Franey >These comments are generated from the ideas presented at the symposium >entitled: >TCP/IP & OSI, The Interoperability Challenge, Washington, D.C., >May 21-23 1990 hosted by the Technology Transfer Institute, Santa Monica, CA. >These ideas are not mine and I am not speaking for my company in any respect. -------------------------------------------------------------------------------- >From: att!lzaz!jer >I support your suggestion. >joe ritacco -------------------------------------------------------------------------------- >From: uunet!sunkist.West.Sun.COM!gene.saunders (Gene Saunders Sun SE Irvine CA) >I'd suggest the newsgroup be named "comp.protocols.tpicp.to.iso". >More descriptive. >Gene Saunders \ Gene.Saunders@West.Sun.COM \ 72265,23 (CompuServe) >Systems Engineer \ saunders@sunkist (local) \ GSAUNDERS (GEnie) >Sun Microsystems \ #saunder@ProteonW (from Novell) \ ..!sunkist!saunders -------------------------------------------------------------------------------- >From: Amit Parghi <uunet!watmath!watcgl.waterloo.edu!aparghi> >Organization: Computer Graphics Lab, University of Waterloo >Don't use the name "comp.protocols.migrate.to.iso"; since there are >already groups under "comp.protocols.iso", call it >"comp.protocols.iso.migration". >You claim the state-of-the-art engineers in Korea and Japan don't much >care for TCP/IP; instead, they use (I assume) ISO solutions, which I guess >would mean full ISO stacks implemented efficiently across heterogeneous >hardware and physical media. I, for one, would be very interested to hear >of such developments, especially since serious development of efficient and >*useful* implementations in Europe and North America still seems to be some >ways away. Do tell us more. -------------------------------------------------------------------------------- >From: grieve@cos.com (David Grieve) >Organization: Corporation for Open Systems, McLean, VA >I would call it "comp.tcp.osi.transition" >Migration means you go away and then come back. Transition means >you go away for good. >grieve@cos.com OR {uunet, decuac, sun!sundc, hadron, hqda-ai}!cos!grieve >DISCLAIMER: Opinions expressed are not necessarily those of the >Corporation for Open Systems, its members, or any standards body. -------------------------------------------------------------------------------- >From: nsc!asylum.sf.ca.us!karl (Karl Auerbach) > I propose a comp.protocols.migrates.back.to.tcp group to >handle the situation when folks find out that ASN.1 eats all the cpu >cycles and that the word "open" in OSI is something akin to the "open" >invitation that a snake makes to a mouse when the snake opens its >mouth. I think this one was one of the junk mails. But it's funny !! -------------------------------------------------------------------------------
goldstein@carafe.enet.dec.com (Fred R. Goldstein) (06/12/90)
In article <4460@infmx.UUCP>, kwang@infmx.UUCP (Kwang Sung) writes... > Recently, I've proposed a new newsgroup "comp.protocols.migrate.to.iso". >I've received so many responses. Among those, I've selected some best ones, >after I've thrown out some bad ones. I couldn't reply to all of you, since it's >so many. However, I might be able to answer thru this network. I think some >people still don't understand how/where this networking world is going to. >Absolutely, and positively, the world is moving toward one OSI world !!, >even if ISODE has screwed up OSI world a little bit. John Franey, who works for the same company that I do (and I don't think either of us is speaking officially here!) says, >TCP/IP is connectionless, OSI is connection oriented. >Multimedia needs multiple synchronized data circuits. Datagrams can't be lost >or take variously different routes (TCP/IP). That basically is wrong. TCP/IP is connectionless (CL) at the top of the network layer. OSI offers BOTH CO and CL services. DECnet/OSI provies the CLNS, not unlike IP. It also supports the CONS when needed for compatibility, but its "native language" is connectionless. ISO settled the CLNS/CONS war by standardizing both. And multimedia is irrelevant. If anything, the higher speeds possible in CLNS tend to make it more applicable for multimedia than CONS, which is X.25-based. >DEC has commited itself to OSI. All of its customers computer network nodes >are going to be converted to OSI with the release of DECnet Phase V. I work >for DEC but not in communications engineering but please don't think this is a >plug. While I'm not a spokesman, I do work for DEC in communications engineering. We have certainly taken a leading role in OSI. But we are also building TCP/IP. And while Phase V is OSI up to Layer 3 it supports OSI and non-OSI in parallel using a "towers" approach. Most DEC-DEC applications won't use the OSI upper layers. Note that it's the upper layers of OSI that are most controversial. >TCP/IP is NOT popular in Europe. It doesn't surprise me that its not popular >in Korea either. The big reason here is that the PTTs that implement and >use the networks in Europe are connection oriented. They wand Virtual >Circuits. TCP/IP is a datagram network and is not acceptable. The PTTs only own their own internal networks. They provide X.25 subnetworks. In OSI terminology (see ISO8648) ANY public network is just a subnetwork; you may run an NS-protocol on top of it. So you can run OSI-CLNS (ISO8473, found in DECnet/OSI) and be fully OSI compliant, or you can run IP above it and be fully TCP/IP compliant. The PTTs have no say in what you do above their subnet layers. Datagrams are not easy to bill for, so the subnet is still CO. I post this because I don't think the terms of reference of this debate should be based on misinformation, like "OSI is connection-oriented" or "PTTs won't allow TCP/IP to be used in Europe." OSI CLNS + TP4 is not very different (semantically) from IP and TCP. They provide a growth path, especially since IP addressing has obvious limitations. But they also provide coexistence. TCP/IP is a valued technology and while comp.protocols.osi.migration might be a good group to have, extremist anti-TCP/IP rhetoric will make the transition harder and more painful for all involved. --- Fred R. Goldstein goldstein@carafe.enet.dec.com or goldstein@delni.enet.dec.com voice: +1 508 486 7388 opinions are mine alone; sharing requires permission