On Fri, Dec 23, 2011 at 3:06 PM, Tomas Podermanski <tpoder@cis.vutbr.cz> wrote:
Can you be more specific? I can not imagine situation where SLAAC could solves a problem that DHCP would not.
SLAAC is the magic that makes the link-local scope work. I think having a link-local scope is a good thing, so I think I'll keep SLAAC. Now that I'm keeping SLAAC, I think I might as well make it an option for global unicast addressing. DHCPv6, especially on a large scale, does have a cost. A small network doesn't need much of a server, but for a large network the amount of requests can be high. DHCP is also something that isn't trivial to distribute across systems to avoid a single point of failure, there is an entire discussion on the design issues of making a salable DHCP solution, especially if you want more than a generic open pool. I'd say being able to use SLAAC and avoid that complexity is something worth while. RA is much more responsive than DHCP was. When an IPv6 router goes away, hosts can release global addresses for that prefix and fail over gracefully, rather than depending on stale configuration data and blindly sending packets. In the future, we'll likely see RA leveraged to provide better availability than we've seen with IPv4. Then there is the entire issue of someone misconfiguring a DHCP server and having to run around rebooting systems to get them to drop the bogus information (or wait for leases to expire, typically several hours). At least with RA + DHCPv6 you can recover from this in a reasonable amount of time. There are other special case considerations; extensions like privacy addressing kind of become not so private if everything is being logged by a DHCPv6 server. It's legitimate that you might want a network where the anonymity of users is provided. Especially as we continue to see increased requirements for log retention by governments. -- Ray Soucy Epic Communications Specialist Phone: +1 (207) 561-3526 Networkmaine, a Unit of the University of Maine System http://www.networkmaine.net/