>keep in mind that OSI is a reference model, and may or may not have anything
>to do with the way things really work. if I am using my word processor, and
>want to open a file located Out There, I perform a couple of actions, each
>of which triggers subroutines in the word processor, which in turn passes
>calls to the OS which in turn performs certain actions, including
>interfacing with network hardware. For purposes of discussion one might say
>that things are being passed up and down the OSI layers, but practically
>speaking, all that is happening is that bits are being pushed and pulled out
>of processor registers.
>
>In terms of OSI, layer 4 is responsible for the call setup and maintaining
>end to end communication. I have yet to figure out what the session layer
>does. I doubt that even Howard can explain it so it makes sense. In fact,
>IIRC, Howard once said that the session layer exists because a committee was
>set up to work on it, and the CCITT of course "had" to incorporate their
>work into the model. ( forgive me if I misremember your comment, Howard.
>It's been a while )
>
>Chuck
Reaching back into my non-silicon memory, the original proposals for
OSI Transport included an advanced version that did, indeed, have the
concept of some record-level integrity. There was, however, an
existing CCITT standard (and committee) for, IIRC, Teletext, which
had an existing standard for record delimiting, the push function,
etc. It was that committee that claimed it had already developed a
solution.,
OSI Transport Protocol, Class 4, the version standardized with the
greatest functionality, does NOT have the graceful close (i.e., PUSH)
that is in TCP. Graceful close is architecturally an OSI session,
not transaction, feature. A reliable transport service ensures
delivery of bytes but not records. TCP has _some_ functions
associated with OSI Layer 5.
At a slightly less serious level, from the days I actively worked
with OSI protocols:
The Seven Deadly Layers
By Howard C. Berkowitz
Among the most frequent questions I'm asked in OSI teaching
is, "Do I need to know what all the layers do?" This is especially
true of management
audiences, who "need to know" the power centers. (They may not kno what a
layer is, but they know there are seven of them and they don't want a
single one to go unsupervised.)
Over the years, I have found a useful analogy. Educational
theory suggests we should start with something that the student knows
and build from there. Therefore, I ask management audiences to
reflect not on theoretical network architecture, but on sin.
Specifically, I ask them to consider the Seven
Deadly Sins (Note 0).
These sins have definite relevance to the OSI Reference Model. The "most
popular" deadly sins are analogies for the layers most important for
non-developers to know about. Audiences think of sins in a fairly consistent
way.
Approximately 75 percent immediately think of Lust. Lust, clearly,
relates directlyto the Physical Layer. It is essential to be aware of
the function of the Lust Layer, for that defines how to "plug in."
(Note 1)
Most of the remaining audience split between Avarice and Gluttony. These
also are important in OSI. Avarice, or Greed, is often realized
as the Bottom Line in business. One is closer to understanding the
Tao of OSI when one realizes that it places the Bottom Line (i.e.
what OSI does for real user applications programs) on Top. The top of
the Avarice Layer is the Service Access Point to the Application, or
Avarice, Layer. ([Note 2)
Those members of the audience who thought first of Gluttony also have some
understanding of OSI. Gluttony deals with establishing a relationship
between a mouth entity and a food entity. Network deals with the next
course while Transport deals with the end goal of dessert. Users
really need to know the functions of Application, Transport/Network
(as the distinction blurs here), and Physical. They often also need
to know the characteristics of the data link layer.
Since Data Link has to deal with collisions, master/slave
relationships, etc., it may correspond to the sin of Anger. I tend
to associate the sin of Pride with the Presentation Layer, on the
grounds that Presentation is rather prideful to think that it
justifies its own layer.
There is always one in the audience, however, who thinks of Sloth. Sloth is a
difficult sin. How does one confess it? "Bless me, I have slothed?"
"Forgive me for committing sloth?" How can I commit not doing
something? Since Sloth is a sin we really have trouble talking
about,and involves not doing useful things, it is a relevant analogy
to the Session Layer.
Both Sloth and Session are needed for theological completeness, but
their relevance to the ordinary sinner or the OSI user is fairly
limited. [Note 4]
If we were to redesign the OSI Reference Model today, its exact number
of layers would be controversial, but it would almost certaily not be
seven. The flight from sevenness is seen in Internet usage and in
the more obscure (but useful) ISO Technical Report 10000 taxonomy for
defining
protocol stacks (i.e., International Standardized Protocols). Internet
practice does use layered protocol stacks, but these rarely have seven
layers (Network File System is an exception). Typically, an application
service such as File Transfer Protocol (FTP) includes the functions of
the top three OSI layers, and in turn layers on top of TCP, a Transport
layer protocol. TCP, in turn, layers on top of IP, a network layer
protocol. IP runs over a generic Interface or Hardware layer that
include Data Link and Physical layers.
In like manner, OSI stack specifications have a top A- or B-specification
that defines a particular set of Application, Presentation, and Session
protocols and their options. The A- or B-profile runs over a lower layer
stack definition, the first letters of which identify the transport and
network protocols in use, followed by numbers referring to the data
link and physical layers. While little known outside OSI circles,
this ISO 10000 methodology is useful as a notation for many multilayerd
protocol stacks. (Note
3) Research funding is being sought to evaluate the possible
relevance of the Seven
Dwarves to OSI.
(Note 0) Just as there are many theological interpretations in
religion, this is not the only interpretation of "sevenness" and OSI
layering. David Piscitello independently presented a different mapping
of sins to layers in the early 1980s. A Mark Russell presentation gave
the author insight on the meaning of confessing sloth.
(Note 1) When presenting these analogies at an IEEE conference in New
York, a woman's clear voice rang out from the back of the room.
"Well, I'm glad SOME
standards body is defining how to plug in things correctly. God
knows most male engineers don't understand that at all."
(Note 2) This part of the analogy can continue into Application
ServiceElements: ACSE (the Avarice Control Service Element), ROSE (the Remote
Organization Submission Element), etc.
(Note 3) After their first reading of Presentation Context Negotiation and
ASN.1 Basic Encoding Rules, some nominate the sin of Pride as the proper
analogy for the Presentation Layer.
(Note 4) Session actually does useful things. The widely used Remote
Procedure Call (RPC) protocol, which underlies NFS, is a pure session
layer protocol. NetBIOS, as distinct from NetBEUI, also essentially
is a session layer protocol.
Message Posted at:
http://www.groupstudy.com/form/read.php?f=7&i=28783&t=28378
--------------------------------------------------
FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html
Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
