<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
<head>
<meta content="text/html;charset=ISO-8859-15"
http-equiv="Content-Type">
<title></title>
</head>
<body bgcolor="#ffffff" text="#000000">
I had not intended to start a religious argument here when I said the
DHCP server ARPs to find out if the address it's going to assign is
already in use.<br>
<br>
John Summerfield is correct that the specification calls for an ICMP
echo request to be sent.<br>
<br>
Mogens is correct that the ICMP echo is the mechanism used to generate
the ARP request and subsequently verify the existence or absence of the
address on the segment.<br>
<br>
ARP is the mechanism for providing an association between the physical
link layer and the logical link layer entities. IP addresses are
logical link layer entities and ethernet (MAC) addresses are physical
link layer entities.<br>
<br>
ICMP may not <i>depend on</i> <i>knowing</i> how the traffic is
transported but it does depend on knowing that it can be transported.
That's what address resolution is all about. If the logical address
(IP) cannot be resolved to a MAC address it cannot be transported.<br>
<br>
Merry Christmas, Happy Hanukah, and a prosperous New Year to you all.<br>
<br>
--jc<br>
<br>
Mogens Kjaer wrote:<br>
<blockquote cite="mid476B5262.80106@crc.dk" type="cite">John
Summerfield wrote:
<br>
..
<br>
<blockquote type="cite">ICMP doesn't depend on knowing how IP traffic
is transported.
<br>
</blockquote>
<br>
If I ping (send ICMP echo request) a non-existing, local IP
<br>
address, the first thing my machine does is sending an
<br>
arp who-has on the wire. No ICMP package is sent before
<br>
my machine receives the is-at answer.
<br>
<br>
Mogens
<br>
<br>
</blockquote>
</body>
</html>