to be
unique, existing MAC addresses aren’t guaranteed to be unique [7, 3]. There are known instances
where network interface cards (NICs) have been issued with duplicate MAC addresses. Also not all
hardware has an IEEE MAC address, as it may be using different wireless technologies.
• It would require a change in existing software, as often 32-bit addresses are used.
• Interworking between fixed and ad-hoc networking would be complicated
• The fact that the identity of the node can be determined from its address raises privacy concerns.
• The larger the address used, the greater the per-packet overhead. This becomes more of an issue in
source routing protocols where a list of addresses is placed in each packet.
Mobile-IP based solutions (e.g. MIPMANET [6]) typically require nodes to have a permanent home
address, which is used as the unique address within the ad-hoc network. All existing hardware then
requires a permanent address, and the increased address size that would be required (probably requiring
the migration to IPv6 due to the shortage in IPv4 addresses) would require software changes, and create
extra overheads. It would also create new issues if the ad-hoc network was to become connected to the
global network, with the selection of a care-of address either through the selection of an appropriate
foreign agent (as would be the case if IPv4 were used), or the configuration of a local care-of address.
For example with MIPv6 the mechanism for auto-configuring a care-of address [7] still requires
Duplicate Address Detection (DAD), even though the address has the MAC address embedded in it.
Performing this DAD in an ad-hoc network is not a trivial matter, and a new scheme is required to
manage this autoconfiguration of addresses.
If we discount the use of a static address then a scheme is required to allow nodes to autoconfigure an
address within an ad-hoc network. Current solutions typically have limitations, working only in certain
scenarios. The scheme we present provides an ad-hoc autoconfiguration mechanism, that isn’t reliant
on the MAC address and works in an ad-hoc environment. It is based on the use of IPv4 addresses, but
this could easily be extended to use IPv6 addresses. It aims to provide a solution that minimises both
traffic and time requirements for configuration and management. We also briefly examine how these
networks could be connected to the fixed network.
2 Related Work
Perkins et al [1] have proposed a simple Duplicate Address Detection (DAD) based scheme whereby
nodes choose a tentative address and send a request to this address. The absence of a reply is assumed
to indicate that this address is free for use. However there are a few limitations to this approach.
• In the event of a partitioned network, or the coalescing of two http://www.51lunwen.org/jisuanjiyingtyu/2011/0526/lw201105261601196514.htmlnetworks, the address would not be
guaranteed to be unique and the procedure would have to be repeated again after merging.
• No mechanism is suggested to detect the merging of networks, and even if such a mechanism were
in place, every single node in both networks would have to redo this DAD test, potentially causing a
disruptive broadcast storm. Whilst this problem is recognised, no solution is suggested.
• Timeout-based approaches suggest that the interval required for a packet to successfully traverse the
entire network can be calculated, whereas the nature of ad-hoc networks makes this difficult.
• Joining the network takes a long period, as a number of
