Short version:    As far as I know, the text Xiaohu points to is
                  not relevant to the question of whether a single
                  ILNP FQDN can lead to multiple unique hosts, each
                  of which has a single Identifier and potentially
                  multiple Locators.

Hi Xiaohu,

You wrote:

>> From what I understand about ILNP, Tony is right to say the FQDN is
>> not the Identifier.  A FQDN lookup can return multiple Identifiers,
>> each for a different host.
> No. if you see the LP record and its usage (see the following quotes) in
> ILNP, you will infer that the multiple identifiers associated with a given
> FQDN belong to a single host, rather than different hosts.
>    "...In the new scheme, site multi-homing works in a similar manner,
>    with nodes having one Locator for each upstream connection to
>    the Internet.  To avoid a DNS Update burst when a site or
>    subnetwork moves location, a DNS record optimisation is
>    possible.  This would change the number of DNS Updates required
>    from Order(number of nodes at the site/subnetwork that moved)
>    to Order(1). [ILNP-DNS] "

OK, this is from the latest ID, not yet at the IETF site:

Here is my understanding of how ILNP works, or should work.  I am
busy with RANGER and won't try to re-read the ILNP material now.

If a host has an Identifier ZZZZ ZZZZ (64 bits) and has two Locators:

    AAAA AAAA (64 bits, from ISP-A)

    BBBB BBBB (64 bits, from ISP-B)

then the host has, in conventional terms, two IP addresses, each one
being an "upstream connection to the Internet."

    AAAA AAAA ZZZZ ZZZZ  128 bit address, with packets arriving via

    BBBB BBBB ZZZZ ZZZZ  128 bit address, with packets arriving via

There may similarly be another host in the same end-user network,
with Identifier YYYY YYYY.  Its two (in conventional terms) IPv6
addresses would be:


The paragraph you quoted refers to a mechanism by which, if this
end-user network stops using ISP-B, but instead starts using ISP-C
(which gives all the hosts in this network a new Locator CCCC CCCC,
then there will be a protocol for the end-user network to tell the
DNS server, with a simple command, that all the BBBB BBBB xxxx xxxx
entries should be deleted, one for each host, and be replaced by CCCC
CCCC xxxx xxxx entries.  This is more concise then having to tell the
DNS server:

   For Identifier YYYY YYYY, delete Locator BBBB BBBB
   For Identifier YYYY YYYY, add Locator    CCCC CCCC

   For Identifier ZZZZ ZZZZ, delete Locator BBBB BBBB
   For Identifier ZZZZ ZZZZ, add Locator    CCCC CCCC

etc. for each of the hosts in this network.

As best I understand it, the implementation if this in the zone files
is something like specifying, for each host in this end-user network,
two place holders for Locators, rather than the actual locators.

If all the hosts, YYYY YYYY, ZZZZ ZZZZ etc. each have the same two
place holders:  JJ and KK, and JJ was elsewhere defined to be AAAA
AAAA and KK was elsewhere defined as BBBB BBBB, then it will be easy
for the end-user network to tell the DNS server that all its hosts
are no longer using BBBB BBBB as their second locator, but are using
CCCC CCCC instead, simply by sending a single command to redefine KK.

> 2.3 "LP" Resource Record
>    (Description of "LP" record use goes here.)
>    An I record has the following logical components:
>         <domain-name>  IN  I  <preference>   <domain-name-2>

This is from:

I think these last two lines are incorrect.  I think Ran did a copy
and paste operation from the previous section, and forgot to modify
them to be the LP record.

The LP record is explained in page 8 of draft-rja-ilnp-intro-03.
This LP record and the paragraph you quoted above do not seem to have
anything to do with the question:

   How does ILNP do the equivalent of "round-robin DNS", in which
   a single FQDN leads to multiple hosts - and in ILNP, how this
   works with each host having multiple locators.

My understanding is that in ILNP a FQDN can return, when looked up in
the DNS, one, two or more Identifiers, each with one or more
Locators.  Since each Identifier uniquely identifies a host which
could be anywhere at all, each such host (each such Identifier) could
have one or more Locators, independently of whatever the other hosts
(Identifiers) have.

So looking up the FQDN "" may return:

   Identifier  MMMM MMMM   Locator(s)  AAAA AAAA
                                       DDDD DDDD

   Identifier  NNNN NNNN   Locator(s)  AAAA AAAA
                                       DDDD DDDD

   Identifier  OOOO OOOO   Locator(s)  CCCC CCCC

   Identifier  PPPP PPPP   Locator(s)  FFFF FFFF
                                       GGGG GGGG

Each character represents 8 bits.  Locators go into the most
significant 64 bits of the IPv6 address and Identifiers go into the
least significant 64 bits.

In this case, the first two would be hosts in the one physical
network, which is connected to the Net via two ISPs, one of which
provides a PA prefix AAAA AAAA /64 and the other which provides a PA
prefix DDDD DDDD /64

The third host, the one with Identifier OOOO OOOO, would be in
another physical network somewhere, which is served by a single ISP,
which provides that network with a PA prefix CCCC CCCC /64.

The fourth host, with Identifier PPPP PPPP would be in yet another
network somewhere, served by two other ISPs.

Exactly how ILNP does this, I am not sure.  Maybe it doesn't!  This
is my understanding of how ILNP could and should work.

I am pretty sure my naming model diagram for ILNP is correct:

          Role             Level            Some CEE architectures
                                            such as ILNP
          Text name  <---- FQDN
          Identifier <----           IIII IIII } Combined into
          Locator    <---- LLLL LLLL           } one IPv6 address

Ideally, Ran would confirm this.

  - Robin

rrg mailing list

Reply via email to