I am not deny there will be waste. My point is only it is not as much waste as 
1/2^n. However, giving /48 to every subscriber, for me, is much bigger waste 
and luxury. You are assuming every subscriber has as many as than 2^16 subnets? 
Currently, there may not be more than 16 devices in an average subscriber’s own 
network.  On other side, why subscriber may want more than one subnets? One 
reasonable reason is to organize different traffics or applications in 
different subnets. So, these kind of traffic differentiation is also meaningful 
semantic for providers. If the provider has differentiate these traffics in 
higher bits of prefixes using semantic prefix, the smaller prefix subscriber 
are needed. Maybe /58, or /60 is enough for subscribers, if the providers has 
well separate the traffic for them by assign multiple prefix regarding to 
different semantic?

Cheers,

Sheng

From: Owen DeLong [mailto:[email protected]]
Sent: Friday, May 31, 2013 12:08 AM
To: Sheng Jiang
Cc: Lorenzo Colitti; Tim Chown; <[email protected]>; 
[email protected]; [email protected]
Subject: Re: [v6ops] Could IPv6 address be more than 
locator?//draft-jiang-v6ops-semantic-prefix-03

>>However, it is not necessary as worse as 2^N times. For example, it there are 
>>2 bits to separate different use types (say 4 different types), it actually 
>>only separate use address spaces into four different spaces. It does not 
>>limit the address space to be 1/4 of original space.
How is that different from saying "by adding two bits of semantics in the 
prefix, the network will use 4 times the address space than it would otherwise"?
No. This is very different. Putting the example into numbers may be more 
intuitionistic. Say an ISP has 4 million subscribers, it needs 4 million /56 
(assuming every user get a /56). By separating them into 4 different types, the 
address consumption is still 4 million /56 if the separation is exactly even. 
However, the more
Not a great assumption... They should need 4 million or more /48s since every 
subscriber is at least one end site and every subscriber end site should 
receive a /48.

Assuming that you will get 4 million subscribers that conveniently divide into 
buckets of 1 million per bucket is absurd. More likely, you'll get 500,000, 
750,000, 2,000,000, and 750,000, or other similarly skewed distribution. It 
might even be 3,500,000, 125,000, 125,000, 250,000.
address may need when the separation is not even. For example, if the biggest 
user type has 2 million users, then the total address space may become 8 
million /56 – two times of original. It comes from align. The increased 
semantics bit are also increasing the address space although not increasing 
linearly. So, at the end, it is not totally waste.
If you get extraordinarily lucky, it's no waste. Otherwise, it's at least 50% 
waste and can easily reach 75% waste (4x space utilization, as Lorenzo said).

For example, if you have 4,000,000 end sites and you get as little as 2,097,153 
subscribers in one of the buckets, you have to go to 4x your address space to 
preserve the semantics.

Owen

Best regards,
Sheng
From: Lorenzo Colitti [mailto:[email protected]]
Sent: Thursday, May 30, 2013 3:19 PM
To: Sheng Jiang
Cc: Tim Chown; Owen DeLong; <[email protected]<mailto:[email protected]>>; 
[email protected]<mailto:[email protected]>;
 [email protected]<mailto:[email protected]>
Subject: Re: [v6ops] Could IPv6 address be more than 
locator?//draft-jiang-v6ops-semantic-prefix-03

On Thu, May 30, 2013 at 4:13 PM, Sheng Jiang 
<[email protected]<mailto:[email protected]>> wrote:
Yes, there is no intension to change ARIN’s policy at all. ARIN should remain 
the current policy of assign IPv6 address block. But the network providers, who 
has already get address block, can choose to use the addresses with certain 
semantics. And no one, including ARIN can stop this.

Agreed, but even for network providers that already have blocks, this will be 
an issue if they ever need another block. But I do think you should write this 
in the draft.

However, it is not necessary as worse as 2^N times. For example, it there are 2 
bits to separate different use types (say 4 different types), it actually only 
separate use address spaces into four different spaces. It does not limit the 
address space to be 1/4 of original space.

How is that different from saying "by adding two bits of semantics in the 
prefix, the network will use 4 times the address space than it would otherwise"?
--------------------------------------------------------------------
IETF IPv6 working group mailing list
[email protected]
Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------

Reply via email to