Hi Andrew,

4 or 6 subnet bits are not sufficient for most uses.  It is expected
that these addresses will be used for internal numbering within an
enterprise, so we should, at minimum, leave room for a 16-bit subnet
ID.  This would allow the same subnet numbers to be used for global
and local subnets.

So, it doesn't seem likely that we can auto-generate globally
unique addresses within the FECO::/10 space, unless we can determine
a mechanism for routers to obtain globally unique identifiers that
are no longer than 38-bits.

One or more people have made proposals on the list for mechanisms to
generate mostly-unique addresses in the FECO::/10 range, based on
hashes of the EUI-64, among other things...  Hopefully one of those
people will fully document such a proposal in an I-D, so that we can
discuss it in detail.

Margaret

At 05:41 PM 11/26/2002 +1100, Andrew White wrote:
Some thoughts:

As a method of doing globally unique site local addressing:

Assuming aggregability is not an issue within a 'site' sized network,
consider generating site local subnet identifiers at the router, based on
IEEE EUI-48 identifiers (such as MAC addresses).

For example, generate as fec0::/12:

12 bits: fef
48 bits: MAC
 4 bits: 0 or subnets

or, if we don't want them in fec0::/10

10 bits: fe0
48 bits: MAC
 6 bits: 0 or subnets

The '0 or subnets' is to allow for the possibility of choosing one EUI-48 on
a router and using that to allocate all appropriate subnets.

By piggybacking on the existing registration scheme, we generate "unique"
site-local subnet ids at the router without needing external registration or
administration.

Despite the zero-config nature of this, administration on the router is
still necessary, both to enable this mode (probably don't want site-local
behaviour enabled by default) and to determine whether a router is
authoritative for the link.  An administrator may wish to configure a
multi-router link with the subnet prefix of only one router.

An internet-draft describing this in more detail is written and will be
submitted in the next day or so.  Comments welcome.


Another comment on uniqueness:

Under IPv6, even an ambiguous prefix is likely to not resolve to an address
because of the MAC generated machine id, so the likelihood of collision is
lower than might be expected from the prefix.


Final thought:

Whatever the outcome of the site local discussions, renumbering will remain
a serious problem under IPv6, that needs to be considered.  Making
renumbering easier is a hard problem, but a good solution will help reduce a
variety of other problems.

--
Andrew White                [EMAIL PROTECTED]
--------------------------------------------------------------------
IETF IPng Working Group Mailing List
IPng Home Page:                      http://playground.sun.com/ipng
FTP archive:                      ftp://playground.sun.com/pub/ipng
Direct all administrative requests to [EMAIL PROTECTED]
--------------------------------------------------------------------

--------------------------------------------------------------------
IETF IPng Working Group Mailing List
IPng Home Page:                      http://playground.sun.com/ipng
FTP archive:                      ftp://playground.sun.com/pub/ipng
Direct all administrative requests to [EMAIL PROTECTED]
--------------------------------------------------------------------

Reply via email to