Questions in line...
- Ralph
On Aug 23, 2006, at 3:54 AM, Syam Madanapalli wrote:
Hi,
On 8/23/06, Ole Troan <[EMAIL PROTECTED]> wrote:
I don't understand the rationale for this work either.
the first PD proposal (by Brian Haberman) was indeed based on using
ICMP as transport. separate message types instead of
piggy-backing on RS/RA though. as we continued to develop that
mechanism we realised that we were pretty much reinventing the DHCP
protocol machine. that realisation led to specifying it as an DHCP
option instead of as a new protocol.
Currently DHCP mechanism works only between routers whereas this
new mechanism works for end hosts. This draft just introduces a flag
in Prefix Information Flag (PIO) to indicate that the prefix is
unique for
the device. And another flag in RS to inform the routers that the
host is
looking for a unique prefix.
Perhaps we need to back up and understand the application better. In
DHCPv6 PD, we labeled the client node as "requesting router" because
PD seems only applicable in the case where the DHCPv6 client is
acting as a router between some links (to which the delegated
prefixes will be assigned) and the interface through which the
delegated prefix was assigned. Is the ICMP PD proposal addressing
some different scenario? If so, where can I learn about it?
Currently we use RS/RA to discover the prefixes available on the link
(shared prefixes), I think it is natural to extend this for hosts
to get unique
prefixes.
Where are these unique prefixes being used? On the link between the
aggregation device and the CPE or between the CPE and the subscriber
CPEs in the diagram below?
______________________ \
/ \ \
| ISP core network | \
\__________ ___________/ |
| |
+-------+-------+ |
| Aggregation | | ISP
| device | | network
| (delegating | |
| router) | |
+-------+-------+ |
| /
|DSL to subscriber /
|premises /
|
+------+------+ \
| CPE | \
| (requesting | \
| router) | |
+----+---+----+ |
| | | Subscriber
---+-------------+-----+- -+-----+-------------+--- | network
| | | | |
+----+-----+ +-----+----+ +----+-----+ +-----+----+ |
|Subscriber| |Subscriber| |Subscriber| |Subscriber| /
| PC | | PC | | PC | | PC | /
+----------+ +----------+ +----------+ +----------+ /
Also, the subnet model that NetLMM WG wants to choose is to have
a unique prefix for each MN. These extensions for RS/RA allows
NetLMM hosts to be able to get unique prefixes in a better compared
to current RS/RA messages.
a technical comment on your proposal; how do you plan to support
shared links with multiple requesting routers? RAs are typically
multicast.
Typically RAs are sent in unicast in response to an RS. When the
router
receives an RS with the prefix delegation request flag, the router
must
send an unicast RA.
-Syam
/ot
--------------------------------------------------------------------
IETF IPv6 working group mailing list
ipv6@ietf.org
Administrative Requests: https://www1.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------
--------------------------------------------------------------------
IETF IPv6 working group mailing list
ipv6@ietf.org
Administrative Requests: https://www1.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------
