I love to argue, even with myself. One of the problems, ignoring your completely correct statement that TCP/IP has never and will never be OSI compliant, is that the OSI reference model itself has evolved since the original 1984 publication of document ISO 7498 without appendices. Around 1988 or so, the view of the upper layers changed considerably, and there was movement to an object-oriented model among functions in layer 5-7 rather than strict layering. The network layer was also revised so it overlapped into the data link layer, and cleanly dealt with such things as ARP. There were supplemental architecture documents on routing and on management that cleared up a number of the confusions that come from the simplistic model that people are given. By simplistic model, I will merely point out that there is a time in life where it is useful to tell children "the daddy has the seed and gives the seed to the mommy."
That model doesn't hold up in adolescence, and the classic 7 layer model doesn't hold up in real networking. Unfortunately, some of the more important concepts that have held up are abstract -- the idea of services versus protocols, the idea of service information encapsulated in protocol data units, a convenient notation for showing the hierarchical layering of protocols, etc., are rarely taught outside graduate-level computer science programs. >I once had an interesting, if heated argument with someone off list about >this. IIRC, I was told by that person that Cisco, in its current CCNP study >materials, is saying just that - that something operates at the OSI layer >above which it functions. I.e. if a routing protocol uses an IP protocol >number, then it is operating at transport layer. The fallacy in this argument is it assumes that management and control follows exactly the same stack as does user data. >Since BGP uses TCP port >179, it is operating at the session layer, along with RIP, which uses UDP >port 520. ( BTW, I have also read in a reputable source that UDP is >application layer because it is not reliable, and therefore cannot be >transport layer, and there is no place else it really fits ) Your source, by its own logic, is an application layer function, because it is not reliable. > >I recognize that Cisco just LOVES the OSI model in the lower level >certifications, but the fact is that in terms of how things work it is crap, >and tends to cause more confusion and add no value. > >Every vendor of content switches is calling them layer 4-7 switches. what >kind of crap is that? Arguing with myself a bit, I can make a reasonable argument for layer 4 switching, as in TCP load distribution in schemes such as NAT with port translation and load sharing. Using layer 7 (e.g., URL) to do additional resolution of the destination IP address could be called relaying, as it is a little more direct than a directory lookup. NFS portmapper is sort of a host-based switching function. Layer 6 switching makes no sense. I suppose something like a RPC load distributor could do layer 5 switching, but the reality is that distinct layer 5 and 6 protocols are very rare in IP practice. >I dare anyone to justify switching as a layer 5 or a layer 6 activity. Yet >there it is. Also, to judge from what content switches do, the marketers are >saying the OSI layer 7 is user application, not a service application, >something Howard takes great pain to differentiate in his writings on the >subject, again IIRC. > >TCP/IP is NOT OSI compliant, never has been, never will be. OSI is a >reference model, and not necessarily related to anything in real life. > >End of rant. > >Chuck > > > >-----Original Message----- >From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]]On Behalf Of >Jose Luis De Abreu >Sent: Thursday, December 13, 2001 12:25 PM >To: [EMAIL PROTECTED] >Subject: Routing protocols [7:29139] > > >Just an open question ? > >We read, learn and teach Routing protocols are at the >NETWORK layer of the famous OSI model... > >But they have PROTOCOLS NUMBERS - TRANSPORT LAYER(such >as IGRP protocol 9, EIGRP protocol 88 and OSPF >protocol 89)and APPLICATION PORTS values - APPLICATION >LAYER (RIP uses port 520 and BGP4 uses port 179) >indicating they work in the upper layers and not in >the network layer, although the result is shown int >the NETWORK layer... > >So may question is... > >Do they really operate at LAYER 3 ? > >Warm regards, > >Jose Luis De Abreu > > > > > >______________________________________________________ >Send your holiday cheer with http://greetings.yahoo.ca Message Posted at: http://www.groupstudy.com/form/read.php?f=7&i=29188&t=29139 -------------------------------------------------- FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
