> Date: Thu, 30 Sep 1999 19:25:10 +0100 (BST)
> From: Dirk Koopman <[EMAIL PROTECTED]>
> Subject: Re: AX.25 Layer 1
> [...]
> Not where he did. Since the bit stuffing is an integral part of HDLC (and is
> implemented by the SCC chip together with flag detection and CRC checking), I
> declare it to be part of layer 1, ax25 is layer 2, the two netrom layers 3
> and 4 respectively.
I suppose we are all free to declare anything we want, but the use of
common terminology has been know to aid communications.
In general, one ought to be cautious about classifying protocols based on
_implementation_ rather than on _function_. If I implement flag detection,
bit stuffing and CRC calculation on my PIC processor rather than an
outboard chip, I probably ought not reclassify HDLC into another protocol
layer.
Without resorting to a textbook, the physical layer is responsible for
encoding data onto the physical medium, the link layer is responsible for
reliable communications across a link between adjacent nodes, and the
network layer is responsible for communications between end nodes, which
may not be adjacent.
The boundaries between these layers are at times fuzzy; some protocols don't
map nicely into the standard seven layer model, plus some protocols
appear a bit confused about the boundaries between protocol layers.
> There are several that reckon ax25 is actually layer 3 with hdlc layer 2, but
> they also believe that Linus rules the universe as opposed to merely being
> immortal.
(Note that the X.25 spec last used the "layer 2" and "layer 3" terminology
in the 1976 spec and used "link layer" and "packet layer" in the 1980 and
later specs to reduce layering confusion.)
X.25 describes two protocols, a link layer protocol that is pretty close
to HDLC that is responsible for communications between two adjacent
nodes, (actually between the customer node and the network node), and a
packet layer protocol that deals with end-to-end communications (but,
due to the architecture of X.25 networks is not exactly an end-to-end
protocol...).
AX.25 munged link layer and end-to-end functions into one protocol,
(although the document contains hints that this was intended to be only
an interim solution). (Jack Taylor might wish to express an opinion
here...) The link layer portion of AX.25 includes the HDLC framing
(flags and bit stuffing), the CRC and probably the address extension bit.
The end-to-end portion of AX.25 includes the addresses (well, most of the
time; the "via" addresses obviously have end-to-end significance, but
the "to" and "from" are in some sense used by both the link layer and
the end-to-end layer) and the control byte and the PID. So, given that
link-layer functions and end-to-end functions are munged together, it
is easy to see how people might differ in classifying AX.25.
-tjs
