Path: utzoo!mnetor!uunet!husc6!linus!philabs!ttidca!retix!erik From: erik@retix.retix.COM (Erik Forsberg) Newsgroups: comp.protocols.tcp-ip Subject: Re: OSI does not mean X.25 Message-ID: <330@retix.retix.COM> Date: 12 Apr 88 07:25:12 GMT References: <76.008873@adam.DG.COM> <1157@ttds.UUCP> Reply-To: erik@retix.UUCP (Erik Forsberg) Organization: Retix, Santa Monica CA Lines: 68 In article <1157@ttds.UUCP> rajaei@ttds.UUCP (Hassan Rajaei) writes: >I am really glad you mentioned this. There has been a confusion between >OSI model and X.25 protocol for a long time just because X.25 was the only >available implementation of OSI. > >The OSI model is so general that you may do any thing with it (except the >overhead!). If there is not an standard protocol available for your need >within the model, that doesn't mean the model itself is incapabel of doing >that. In spite of many standard protocols available for OSI at present >time, I believe we need many new ones in future even for the low layers >like physical, link and network. > >The existing standars for low layers are incapable of handling the ultra >super speed networks of the future (FDDI can handle just 150 Mbps). The >same is true with X.25 and its IP X.75 which are not only limited by speed >but rather make the network very vulnerable because of their connection- >oriented behaviour throughout the network (internetworks). As Lyman Chapin >said the limitation is not in the model but in the protocols. > >There is much to be done for OSI model to be accepted (or rejected!) world >wide, both with new standard protocols and implementations. > Please make a distinction between the OSI MODEL and the protocols specified by ISO that implements services defined by the ISO model. I don't think you can do anything with the OSI model. Just because you invent your own protocol, which happens to provide some service defined by the OSI model, doesn't really make this new protocol an OSI protocol. There will be just confusion and interoperability problems if every new protocol claims to be an "OSI protocol". Before it could be considered as a protocol to be used to implement a service as defined by OSI, it should become an ISO standard. Otherwise, it's not too useful for the majority of the worlds data communications users. Anyway, there certainly is a place for new protocols for new, higher performing LAN technologies. But, even the existing ISO 8073/8473 protocol combinations are quite performing. (This is the ISO Class 4 Transport protocol operating over a connection-less network service, almost identical with DoD IP). For example, by eliminating overhead imposed by non-perfect hardware, this protocol combination has proven able to have a substained transport layer user data throughput of more than 2000 packets per second (each packet 1024 bytes) which is approximately 16 Megabits/second (this is measured on a VAX 8650). Now, if you add some well-known, supposedly reasonable Ethernet controllers on an otherwise idle Ethernet network, performance drops to a measly 60-180 packets per second, it is my opinion that controller hardware technology, computer buses and software used to interface with the host operating system needs some large improvements. I do not understand why so many believes that X.25 is the only way to implement OSI. It is certainly true that the european continent started work in ISO, specifying the Connection-oriented network service as examplified by X.25, but I think the US has been as successfull in providing equally good protocols when Local Area Networks are the primary technology of interest. Now, there are very reasonable standards in how to inter-connect multiple Local Area Networks using these venerable and perfectly working X.25's as provided by any Public Data Network service provider (in most any country of the world). One of the major problems is that there is no natural way to interoperate between networks using ISO 8473 (IP) or ISO 8208 (X.25) as the network layer. There will always be limitations when such attempts are made (there are several proposals discussed as of this time). -- ---------------------------------------------------------------------------- Erik Forsberg, Retix, 2644 30th Street, Santa Monica CA 90405 (213) 399-2200 UUCP: {cepu,ttidca,rutgers,oliveb}!retix!erik, erik@retix.com