The topic that will never die... :-)
Priscilla's frame oriented definition certainly is operationally
useful, but doesn't quite fit within the formal OSI architecture.
Mike's point is true with respect to formal protocol, but not to
service definition in very formal OSIism. (Incidentally, I do
consider myself a recovering OSIholic. I will never forget the day
when, while shaving, I suddenly realized I completely understood the
abstract definition of the Presentation Service, and almost cut my
throat).
From a formal OSI standpoint, we can speak of a protocol data unit
with a header at layer (N) as an (N)-PDU. The payload of such a
packet is normally a service data unit of layer (M), or (M)-SDU.
For most user plane communications, the payload in an (N)-PDU is at
layer (N+1), such as TCP in IP. The OSI management architecture
annex to the basic reference model distinguishes between layer
management (e.g., keepalives internal to a layer) and system
management (e.g., SNMP or CMIP from the application layer).
A routing protocol packet, therefore, is layer management, with a
(N)-PDU header and a (N)-SDU payload. I realize this gets ugly with
routing protocols such as RIP and BGP that use a transport header,
but, hey, nobody's prefect.
So ARP has a (2)-PDU header and a payload that contains (2)- and
(3)-information. Since the only motivation to have a PDU is to
exchange payload information, I would argue, if I had to put ARP into
a specific layer -- which I think is coercion -- I would call it a
layer management protocol for layer 3.
Again, we are getting into a situation where there is a desire to
coerce things into a simplified version of the OSI model. Real OSI
documents are very careful about the protocol versus service
definition, and indeed you will find separate documents, say, for the
transport service and the (several) transport services. Doing things
this way completely sidesteps the "what layer is this" problem.
>At 06:52 PM 6/14/01, Michael L. Williams wrote:
>>Isn't arp a layer 2 protocol? I realize the goal of ARP is to find a MAC
>>for a given layer 3 address, but the broadcast is done on layer 2, and the
>>remote station responds with layer 2....
>
>That's a good way of thinking about it that I don't think the rest of us
>thought of, Michael.
>
>An ARP frame has no Layer 3, contrary to what so many books claim. Here is
>an ARP frame for reference. Note that there is no IP layer. The ARP header
>references IP addressing information, but there's no IP header.
>
> Flags: 0x00
> Status: 0x00
> Packet Length:64
> Timestamp: 16:10:31.101000 06/14/2001
>Ethernet Header
> Destination: FF:FF:FF:FF:FF:FF Ethernet Broadcast
> Source: 00:00:0E:D5:C7:E7
> Protocol Type:0x0806 IP ARP
>ARP - Address Resolution Protocol
> Hardware: 1 Ethernet (10Mb)
> Protocol: 0x0800 IP
> Hardware Address Length: 6
> Protocol Address Length: 4
> Operation: 1 ARP Request
> Sender Hardware Address: 00:00:0E:D5:C7:E7
> Sender Internet Address: 10.0.0.1
> Target Hardware Address: 00:00:00:00:00:00 (ignored)
> Target Internet Address: 10.0.0.2
>
>Some people might argue that the frame format isn't a good way to
>characterize a layer, but I think it is.
>
>And you make a good point that this frame would not go through a router,
>which is another argument in favor of it being at Layer 2. The frame stays
>on its local segment. It's a broadcast frame and it has no layer 3.
>
>Priscilla
>
>
>>the layer 3 information in the ARP
>>request and response is just encapsulated data at that point...... which I
>>don't think qualifies it as a layer 3 procotol. If that were the case, TCP
>>could be considered a layer 4-7 protocol because all of the data it
>>encapsulates comes from all of the above layers......
>>
> >Mike W.
Message Posted at:
http://www.groupstudy.com/form/read.php?f=7&i=8672&t=8335
--------------------------------------------------
FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html
Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
