Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!usc!sdd.hp.com!think.com!linus!nixbur!nixpbe!peun11!josef From: josef@nixdorf.de (Moellers) Newsgroups: comp.protocols.iso Subject: Re: DESPARATE! Question on ISO/DIS10022 Message-ID: Date: 19 Nov 90 07:52:35 GMT References: Sender: news@nixpbe.nixdorf.de Lines: 108 In enag@ifi.uio.no (Erik Naggum) writes: > Hey, I needed this! The weather is lousy over here! >Hey you, get the idea that we have _layering_, here. It's the NETWORK >LAYER which takes care of such things as routing and all that, >including dialling the stupid telephone number, or instructing the >DATA LINK LAYER to do it, which sends the necessary instructions to >the PHYSICAL LAYER, to be more precise. The PHYSICAL LAYER, however, >doesn't have a clue whether it's transferring Physical Layer Service >Data Units (i.e. _bits_ :-) meaningful to a higher layer as dialling >(such as your modem's NETWORK LAYER), or bits meaningful to a higher >layer as this message. The PHYSICAL LAYER knows about BITS and lines >that are ACTIVE or INACTIVE. That's it. End. Period. No more. I know that the network layer is responsible for routing. But I have never seen a modem which has a NETWORK LAYER built in B-{) But as You say Yourself: The physical layer knows about BITS and lines that are ACTIVE or INACTIVE. Now I ask You: when the physical layer ACTIVATES a line, in the case of V.25bis or X.21, activating a line means dialling. >Dialling doesn't _exist_ at the PHYSICAL LAYER. Aaarrgghh! >If you regard the Seven Holy Layers as the Ten Holy Commandments, you >lose. The ISO RM for OSI is intended to describe the interrelation- >ship of functions, not to cast things in stone. Telephone wires >(PHYSICAL LAYER) require a whole set of OSI layers to be useful to >you, qua PHYSICAL LAYER. Considering dialling, PCM encoding, routing, >charging stuff, the whole song and dance, do you think they crammed >all that into the PHYSICAL LAYER? "How do I specify reverse charging >or authenticate the caller using ISO/DIS 10022?" You DON'T. >WHY is it that layering and hierarchical structures are so inordi- >nately difficult to understand? Could it be that people have the >wrong philosophy of concepts? That would explain why Germans seem to >have the most problems all over the ISO framework. It is exactly layering and hierarchical structures I am thinking about (see below for an explanation)! As I am not part of any official work on ISO frameworks I do not apply You commont to my person. >The answer to your "desperate" question goes like this: CCITT chose to >publish a set of principles to guide choosing the layers, in Appendix >A to CCITT X.200. Go _read_ it. _Don't_ come back until you have >_grasped_ the ideas behind each layer, the choices, etc. Then you >will _understand_ that your question makes NIL sense. >Now get this: Eventually, the physical layer does _everything_. That >does not mean it's meaningful to implement X.400 or X.500 in hardware. > Starts getting colder now B-{) >The considerate answer to your question is: I don't think you have >understood the ISO reference model. Even though some physical layer >somewhere is going to output DTMF codes after having detected the dial >tone (V.25bis) or send bits corresponding to the subscriber number >(X.21), you have a _different_ physical layer for DTMF codes than for >QAM signals, which is yet a different physical layer from that which >eventually churns your bits out on the proverbial wire (fiber, copper, >radio, whatever) in the appropriate form. Just as there are many >different application layers, there are many different physical >layers. Now, luckily (!), we have this fantastic reference model to >guide us in eliminating the specific features of each instance of each >layer, and instead talk about more abstract functions and services >provided up and down the hierarchical model. To make my intentions clearer: Imagine someone (e.g. me) was so stupid as to try and experiment with a truely layered implementation of (part of) the OSI protocol stack. One reason for this is exactly what You describe: to have several physical layer implementations from which to choose, e.g. one that does V.25bis, one that does X.21, etc, and then to choose from these. I do know that the ISO reference model is just a "MODEL", but then I'm not referring to the model but to a (draft international) standard that fits into the model. Take e.g. the "Transport Layer Service Definition" (ISO 8072). It was implemented as System V's TLI and if You compare the standard with the implementation, You'll find the same priomitives in both, e.g. T-CONNECT request -> T_CONN_REQ T-CONNECT indication -> T_CONN_IND T-DISCONNECT request -> T_DISCON_REQ I was thinking about doing the same to the physical layer, i.e. implement ISO DIS 10022 e.g. using STREAMS. The physical layer would be implemented as a driver, the interface to the driver would be message based. The messages between the next layer/module up would be based on the primitives given in said standard. Now, either the state of the physical interface doesn't matter and the physical layer just sends whatever it is told to send and it is up to the "network" to interpret the first x characters as a number, or there has to be a way to tell the physical layer that "what now follows is a number to dial" and "what now follows is the data to transfer". >I thought this was blindingly obvious. Sorry, no! -- ======= | Josef Moellers | c/o Siemens Nixdorf Informatonssysteme AG | | USA: mollers.pad@nixdorf.com | Abt. PXD-S14 | | !USA: mollers.pad@nixdorf.de | Heinz-Nixdorf-Ring |