Date: Sun, 6 Oct 2002 21:40:32 -0700
From: "Michel Py"
<michel(_at_)arneill-py(_dot_)sacramento(_dot_)ca(_dot_)us>
Message-ID: <2B81403386729140A3A899A8B39B046405E352(_at_)server2000>
| I'm having trouble explaining to students why a reference model has
| variable definitions.
What you should be explaining to students is that reference models make no
sense as absolutes.
Models are useful only to analyse some particular situation; if you don't
know what the situation is, they just waste time (as this discussion is
showing).
For most purposes, there are only 3 layers that matter, the layer that is
of immediate interest, the layer below, which is used to implement whatever
is being implemented, and the layer above, which is going to make requests,
and consume whatever can be provided. Occasionally 4 will be useful if
you're concerned about the boundary between 2 layers, and then need the
layers either side of those two.
Then that sequence can be placed into the overall stack as many times as
you happen to need it, and depending upon your magnification level when
you're looking, sometimes what appears as one layer to one observer will
be a whole complex stack of layers to another.
Attempting to give these things absolute numbers, other than for ease of
reference in some particular context is lunacy. Even at what we consider
mostly to be the absolute bottom of the stack, I suspect that if you get
the physists involved, you'll find a whole complex set of stuff going on
down there that could be described by a set of layers.
The OSI model has its uses, when you understand that its purpose was to
break up the whole mass of protocols that they were wanting to create
into something manageable to actually work on. Even then, with hindsight,
it is possible to see that they only just barely managed to get the
net and transport defined with some kind of meaning (those were
the ones whose role and function was pretty much known at the time). The
complexity that would grow into the link level wasn't anticipated, and
the less said about the stuff above transport the better.
If you have need of a model to explain some particular situation, then
create on, just don't try and give that, or any other model, and
particular absolute status. It doesn't work.
kre