Tom,
When a datagram passes through a gateway from the Ethernet to some other network, does the source Ethernet address change to that of the gateway computer or does it remain that of the computer from which the message was originally sent (assuming it originated within the Ethernet). Also, when passing from some other network through an Ethernet in route to the destination, does the gateway of the Ethernet that is being passed through create the Ethernet source and destination addresses, i.e. the gateway being the source and the final Ethernet-connected computer being the destination, whether or not that computer is the ultimate destination? I guess I'm confused on when addresses are created and whether or not they can be modified.
Thanks

You may be confused, but your questions suggest that you aren't very confused. The answer to the second question is "yes" and the answer to the first question is "yes, the address changes." Your second question suggests why this has to be true. A packet may go from a non-ethernet network through a router to an ethernet. The incoming packet would have no ethernet source or destination address.

Tom

So on an Ethernet network, everytime a new computer handles the packet, the source address changes to that computer, the rational being that on an Ethernet computers are supposed to ignore all packets that don't apply? Then, say the computer is a gateway to another network that is not Ethernet and the ultimate IP destination is on that network; then there would be no destination Ethernet address in the packet that arrived since the destination was not on an Ethernet?

Now you sound a bit more confused. What do you mean by saying "everytime a new computer handles the packet" on an ethernet. Packets don't get handled by multiple computers on an ethernet. There is one sender and one receiver. The rest of a packet's lifetime is spent travelling on the wire that runs past all computers. It would, however, make sense to ask something about "everytime a new computer handles the packet" on an internet. In this case, all the "new computer"s would be routers needed to transfer the packet from one physical network (ethernet, ring, ...take your pick) to another.

Also you said "there would be no destination Ethernet address in the packet". If the packet is an ethernet packet it has fields for source and destination address (48 bits each) and they must be filled in. The packet can't have "no destination ethernet address."

Tom

OK. So take a computer that is at the gateway between an Ethernet and some other network within the overall Internet. The message is going somewhere on the rest of the Internet (not the Ethernet). Then there is no Ethernet packet in the message sent from the gateway at all? If this is right, then does the gateway receive a message with an Ethernet packet (of which it is the destination), and then send out a message without an Ethernet packet. I think I am making the question way too complicated.

Jan, you're making me confused. But I guess I was already confused.

Jan may be confused (and/or making others confused) but he is very, very close to the issue underlying the problem. So close its hard to answer these questions without just saying the answer.

As for the scenario described in the latest question, a gateway may send a packet without an Ethernet packet. If it is forwarding an IP packet from an ethernet to a token ring it has to take the IP packet out of the ethernet packet and place it in an appropriate token ring packet since those are the only kind of packets acceptable to the token ring.