On Apr 14, 2004, at 10:11 AM, Ralph Droms wrote:
Followup on the meaning of "stateless" - one way to interpret "stateless" in
the context of DHCPv6 is: "does not require the maintenance of any dynamic
state for individual clients" (RFC 3736). The server does, of course,
maintain configuration state and can make decisions about the response sent
to any specific client based on the server's configuration.
Now the words "dynamic state" introduce even more confusion.
A DHCPv6 server that has a pre-computed table to assign parameters
to client is then stateless, the same server that compute this dynamically
(maybe using a simple hash) is still stateless, but if that server decides
to cache this information, it becomes stateful...
And a router sending RAs that keeps track of who it send them to is
by this definition also stateful!
Clearly stateless vs stateful is the wrong way to describe the dichotomy between RA & DHCP based configuration, and this vocabulary should be removed from RFC2462.
(And,
autonomous address assignment isn't "stateless" - the source of the RAs
maintains configuration state, the host maintains state about the addresses
it has selected.) So, "stateless" really is misleading. "Autonomous/managed" or "serverless/server-based" might be more correct...
serverless/serverbased is incorrect too. A router sending and RA in response to an RS
is a server. Autonomous vs managed is IMHO the best terminology.
- Alain.
-------------------------------------------------------------------- IETF IPv6 working group mailing list [EMAIL PROTECTED] Administrative Requests: https://www1.ietf.org/mailman/listinfo/ipv6 --------------------------------------------------------------------
