jj,

> jj wrote:
> 5) I want to have maximum freedom with my /64.
> This is the most important argument, in my mind. I don't
> think any benefit gained from establishing a 64-bit
> boundary can possibly justify loosing the freedom of not
> being confined to that boundary.

You have to be careful using the word "freedom". In this case (prefixes
length >64), freedom is not what you want. Freedom to subnet is what you
want.

There are many forms of freedom.
- The freedom to subnet.
- The freedom of privacy by having changing IP addresses, which requires
a rather large number of bits in order to avoid both collisions and
smart analysis software that identify small ranges.
- The freedom to embed part of a public key in the IID.
- And why not, the freedom to have an IID whose all odd bits are "1"'s,
because it helps synchronizing the frames when running Ethernet over
barb wire.

All these pieces of freedom (purposes) need bits. There are two
approaches to assign bits to a purpose:
1. Reserve some bits to one purpose and some other bits to another
purpose.
2. Allow the same range of bits to be used by several purposes.

Assuming that we are discussing unicast addresses, the current state of
the spec is that bits 0-2 are "001", bits 3-63 are used for routing
(subnetting) and bits 64-127 are used for the IID which includes other
purposes such as the 'u' bit.

What you are favorable to (if I understand correctly) would be to allow
bits 64 to 127 (more likely, bits 64-111) to be used both for routing
(subnetting) and for the IID and other purposes.

1. and 2. can be used simultaneously; the drawback of 1. is that it
leads to bit depletion quickly, and the drawback of 2. is that it makes
purposes that share a range of bits mutually exclusive.


Back to what is being discussed or not:

What is *not* being discussed is your freedom to knowingly violate the
spec, by running your network over barb wire (which requires to change
bits in the IID) or by subnetting above /64 (which requires changing
bits in the IID). This is your very right as long as you understand what
you are doing and even if you don't.

What *is* being discussed is modifying the spec, so it allows you to
legally run your network over barb wire (which requires to change bits
in the IID) or to subnet above /64 (which requires changing bits in the
IID).

The point I will try to make next is why modifying the spec to allow
subnetting above /64 is as irrelevant as modifying it to allow running
your network over barb wire.


Back to your main point:
"I want to have maximum freedom with my /64."

Since we are talking about subnetting, this becomes:
"I want to have maximum freedom to subnet my /64."

This is totally irrelevant. What everyone wants is:
"I want to have maximum freedom to subnet my block."
(The block that has been allocated or assigned to you).

So, let's look at what you are allocated or assigned:

A) You are a small LIR.
If you are a small LIR, you are allocated a /32. Out of this /32, you
allocate two /48s (or 1/32768th, or 0.003%) for /64 ptp links. That
gives you 128K ptp links. What are these ptp links used for? For your
infrastructure, and for customer links that have routed sites. The
breakup of your address space is:
1x     /48 for your corporate network (you are your own customer)
2x     /48 for ptp links
14x    /48 for misc purposes such as /64 dialup / ppp customers
65520x /48 that you assign to customer /48 sites.

Freedom issues here? Not enough ptp links? Make it 8x /48s which gives
you 512K ptp links, that's 0.012% of your address space.


B) You are a large LIR.
Same ratios as A applies.


C) You are a small (read, less than 8K subnets) site.
You are assigned a /48 by your LIR.

Freedom problems connecting 3,000 branch offices with "only" 64K ptp
links/subnets? Come on. And even if it was true, see E).

Will everyone get a /48 if asking for it? Yes. A single v4 IP gives you
a 6to4 /48. Freenet6 gives you a /48. Every LIR or pTLA gives you a /48.
There are proposals such as Tony Hain's that give a /48 to every
backyard-size lot on the planet. We are working on a scheme (see
http://arneill-py.sacramento.ca.us/ipv6mh/geov6.txt ) that gives a /48
to each household out of a single /16.


D) You are a large site.
You are assigned a block relevant to your size, with a reasonable HD
(0.80) that places you in the same situation as C.


E) You did not get a block big enough? Talk to your RIR, this is a
policy matter and not an addressing architecture matter.

This would be Thomas' domain and I won't comment too much on it, what I
hear though is that besides some kind of a policy void for NRENs,
everyone gets the IPv6 address space they need, including enough to use
/64s on ptp links. I have not been following closely the discussions
regarding the HD concerning sites that request more than a /48, but
again this is a policy matter and not an addressing architecture matter.


Freedom to subnet is the topic.

I don't have a freedom problem subnetting with the existing spec. I
certainly would be ok with even more freedom, trouble is that modifying
the spec to grant more subnetting freedom will one day or the other
create an operational issue because the extra subnet bits used by this
un-necessary subnetting freedom will collide with someone that wants
more freedom to do something else with the IID bits.

Michel.


--------------------------------------------------------------------
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