On Jul 6, 10:40am, Ted Lemon wrote:
} 
} > I just reread RFC 2131 (Dynamic Host Configuration Protocol), and
} > I don't see where it specifies this?  It does say that the client
} > identifier must be used to associate a particular lease with a
} > particular client, but it doesn't say that additional information can't
} > be used to determine what address to give a client (I think this would
} > be an administrative issue).
} 
} Technically, I guess you're right.   I could use the host-name field
} as another discriminator in find_lease(), and all that.   But why stop
} there?   Why not allow any arbitrary option to be used as a client

     The only other one that I see being useful is host-name, but that's
just my opinion.

} identifier?   Also, how do I decide when to trust the hostname and
} when to trust the client identifier?   One of the most common reasons

     By what is specified in dhcp.conf.

} why people ask for this feature is that they want to get the same IP
} address for their laptop regardless of what docking station they use,
} or what ethernet card they use.   If I trust the client identifier

     In my case, it was a desktop, but yes that was the general idea.

} over the hostname, this scheme fails miserably.   Without being able
} to support this capability, what you're really asking for is for me to
} let you enter the primary information about a client in the DNS rather
} than in the dhcpd.conf file.   You're also asking me to answer all the

     I would still expect to have to put the necessary information into
dhcpd.conf; although, that would mean duplicating information that may be
in DNS.

} questions that people generate when this completely fails to work!
} :')

     Hmm, okay, I wouldn't wish that on you.  :-)

} > But, I would see the ability to give
} > a fixed-address based on a hostname to be a highly desirable feature.
} 
} That's why the client identifier should be user-settable.   That's why
} it _is_ user-settable in the ISC DHCP client and (I think) the latest

     I agree that everything in the client should be user settable.  The
ISC DHCP client is a very good client (thank-you); but, unfortunately, it
can't be used everywhere, even with source.

} OpenTransport client.   I've heard rumours that Microsoft may follow
} suit, but seen no evidence of it so far.

     Yep, it was a M$ machine.

}-- End of excerpt from Ted Lemon