ietf
[Top] [All Lists]

Re: [Geopriv] Confirmation of GEOPRIV IETF 68 Working Group Hums

2007-04-20 05:33:10
DHCP is not a great choice in a mobile environment and also not when it comes to more complex location representations.

Ciao
Hannes

Brian Rosen wrote:
In the example you gave the Hilton is EXACTLY the network that MUST give you
your location, and Verisign, if they tried, would give a valid, but very
wrong location.

That is the point of using DHCP for location, you need the closest possible
server to get the right location.  You need a server that understands the L2
to which you are connected.  Anything L3 and farther has a big problem of
where, exactly, are you?  The proposals for L7 versions of location
configuration protocols suffer mightily from the problem of figuring out
where you are in the L2.  They have to go to great lengths to determine some
kind of identifier that they can unambiguously use to figure that out.  I
think we have (painfully) figured out a way though that morass that will
work in enough circumstances to be interesting, but it remains hard, very
hard, to identify the L2 when your server is sitting at L7.

So, make sure that when you go to the Hilton that you use it's location
server, or you may have a big problem if you have to make an emergency call
(or even order a pizza).

DHCP is an excellent choice for a location server for networks where the
DHCP infrastructure is present, and can reasonably be upgraded.  The L7 guys
point out, correctly, that that's a tall order in a lot of interesting
networks.  I think that is right.  I do think they believe L7 works on every
network. I'm certain it doesn't.
That's why the compromise of BOTH is probably required.  I know it's the
only way we are going to get anything done in the next year.

Brian

-----Original Message-----
From: Hallam-Baker, Phillip [mailto:pbaker(_at_)verisign(_dot_)com]
Sent: Thursday, April 19, 2007 6:39 PM
To: James M. Polk; Dawson, Martin; John Schnizlein; Andrew Newton
Cc: GEOPRIV WG; ietf(_at_)ietf(_dot_)org; Allison Mankin
Subject: RE: [Geopriv] Confirmation of GEOPRIV IETF 68 Working Group Hums

DHCP is a layer 3 technology that talks directly to layer 2.

This is entirely acceptable, useful and right for NETWORK configuration.
DHCP is an entirely sensible means of obtaining an IP address and
_proposals_ for domain name prefixes and DNS servers.

DHCP should not be used for any other purpose. In particular to make use
of DHCP for application configuration is a layer violation. Layer 7 should
NEVER communicate with Layer 2 directly. When that happens we lose all the
power and flexibility built into the IP stack.


To give a concrete example of the problems caused. I am currently typing
on a VeriSign machine in an office in VeriSign corporate HQ. In that
environment the local DHCP server could provide me with useful and valid
suggestions for all manner of services. But its still the wrong
technology.

The problem is that when I take the machine to the Hilton Garden Inn down
the road where I am staying I explicitly DO NOT want the hotel network to
provide any more than an IP address. I am not going to use their DNS
server and I certainly don't want to make use of any email server, DNS
prefix, GEOPRV or any other application layer feature they might want to
foist onto me.

I am using the Hilton Garden Inn LAN, I am not joining their network. The
machine is remaining on the VeriSign network.


DHCP is a fine technology for the task DHCP is designed to do. It is an
inappropriate technology for application or service configuration. The
proper infrastructure to support those needs is DNS, supplemented if
necessary by HTTP or LDAP backing store (i.e. either discover the services
via DNS directly or use DNS to discover where the directory service is to
be found).

Looking at the history of UPnP and Zero Config it strikes me that
attempting to manage networks through peer to peer broadcast or multicast
have been a bust precisely because of this layer violation.


-----Original Message-----
From: James M. Polk [mailto:jmpolk(_at_)cisco(_dot_)com]
Sent: Thursday, April 19, 2007 5:31 PM
To: Dawson, Martin; John Schnizlein; Andrew Newton
Cc: GEOPRIV WG; ietf(_at_)ietf(_dot_)org; Allison Mankin
Subject: RE: [Geopriv] Confirmation of GEOPRIV IETF 68
Working Group Hums

At 04:20 PM 4/19/2007, Dawson, Martin wrote:
"DHCP is not adequate because it doesn't meet multiple sets of
requirements as documented multiple times ..."
bologna

"documented multiple times" means in individual submissions

of which, zero facts were presented to substantiate

If DHCP were so inadequate, why is the DSL forum now going to
specify it? Why does PacketCable define it?  These were
fairly recent moves...

And, how many times has HELD been presented as if it were a
product of an IETF WG?

James


_______________________________________________
Ietf mailing list
Ietf(_at_)ietf(_dot_)org
https://www1.ietf.org/mailman/listinfo/ietf

_______________________________________________
Ietf mailing list
Ietf(_at_)ietf(_dot_)org
https://www1.ietf.org/mailman/listinfo/ietf


_______________________________________________
Geopriv mailing list
Geopriv(_at_)ietf(_dot_)org
https://www1.ietf.org/mailman/listinfo/geopriv


_______________________________________________
Ietf mailing list
Ietf(_at_)ietf(_dot_)org
https://www1.ietf.org/mailman/listinfo/ietf

<Prev in Thread] Current Thread [Next in Thread>