April 18, 2006

ISO Reference Model - ISORM (as in eyesore)

Recently Ken Camp posted a great overview of ISO Reference Model for Open Systems Interconnect. In this post, I would like to share my understanding of and my thoughts on ISO RM. Many of the points I make here are based on my readings of RFCs by M. A. Padlipsky from early days of “Catenet”. By the way, if you not done so, you will do good by reading him. He is very acerbic and you may find it to difficult but it is worth the effort. Any student of end-to-end argument must do so. By the way I will use Ken’s summary, even though I don’t mean to pick on his writing; it will be easy to refer.

The Internet model and ISO RM couldn’t be any more different. It is ironic that many try to force fit Internet into ISO RM as if it is required for its legitimacy. In fact it should be other way around. Internet has demonstrated that ISO RM has to be changed and changed so drastically, that it shouldn’t be called ISO RM.

Let me start with the Data Link Layer. ISO RM and its realization X.25 did not anticipate multiple access technologies. Both of them visualized only point-to-point links. So MAC protocols like CSMA/CD have no place in ISO RM (not withstanding Ken’s suggestion and the entry in Wikipedia; if you don’t believe me, take a look at the standards). When ISDN Basic Access introduced “passive bus” as the multiple access technology, ITU added it to the physical layer and not LAPD, the link layer.

In my opinion, another significant implication of this assumption is that multiplexing of different flows didn’t take place at the Link Layer. Indeed, when the proponents of Frame Relay encountered cultural opposition because their proposal took advantage of multiplexing at the Link Layer.

For a long time, ISO RM subscribed to link-by-link error correction and not end-to-end. There is a famous paper by Kleinrock where he argues the advantages of the latter. In my opinion three things influenced ISO to change its RM: inherent error-free nature of modern communication systems, success of Internet and Frame Relay. In this respect, ISO RM missed in anticipating an important design principle that has simplified Internet.

Yes, it is beneficial to have a Reference Model. But what is the benefit of maintaining a failed model? Aren’t we better off it assuming the Internet Model?

Posted by aswath at April 18, 2006 08:01 AM
Related Posts Widget for Blogs by LinkWithin
If you do not have an OpenID, then please use www.enthinnai.com/unauopenid/anyblog.

 

Comments



Copyright © 2003-2014 Moca Educational Products.