Brian,

>> The problem with static allocation is that it cannot adapt quickly to 
>> subscriber demand.
> 
> Exactly; so it cannot optimise the number of ports in use.
> 
> What strikes me in this conversation is that all these solutions are a
> tremendous amount of work to deal with legacy equipment, and I have
> to wonder whether that is economically rational. That is for the market
> to decide, but I don't see that the IETF is obliged to solve each
> possible case. In other words I'm not sure where to find the requirements
> analysis that this new burst of activity is based on.

my assumption is that this is done on a CPE, as a part of the existing NAT 
function.
legacy applications exists behind the NAT.

there are multiple implementations of the encapsulation scheme already.

cheers,
Ole

>> Dear Brian,
>> Thank you and in line...
>> 
>> 
>> Best Regards,
>> Tina TSOU
>> http://tinatsou.weebly.com/contact.html
>> 
>> 
>> -----Original Message-----
>> From: Brian E Carpenter [mailto:[email protected]] 
>> Sent: Sunday, August 21, 2011 3:21 PM
>> To: Tina TSOU
>> Cc: Rémi Després; [email protected]; 
>> [email protected]
>> Subject: Re: [Softwires] draft-murakami-softwire-4v6-translation
>> 
>> Hi Tina,
>> 
>> On 2011-08-21 14:08, Tina TSOU wrote:
>> ...
>>> If an average IPv4 user is consuming 200 ports (or whatever
>>> value you prefer to assume) with their favourite p2p app, that
>>> is what sets the number of IPv4 users per shared address. It's
>>> the number of simultaneous ports, not the amount of traffic,
>>> that counts.
>>> [Tina:
>>> These drafts attempt to solve this issue.
>>> https://datatracker.ietf.org/doc/draft-tsou-pcp-natcoord/
>> 
>> This doesn't create new ports; it just helps to allocate
>> them more rationally.
>> [Tina: When I see your comments about consuming 200 ports, I thought about 
>> the motivation of NAT-COORD is to solve the issue of one iTunes, Google Map 
>> application consumes 200~300 ports at a time. I did not think about 
>> Stateful/Stateless at that time.
>> 
>> In the use cases you mentioned above, do you only mean static port 
>> allocation issue with stateless mechanisms? Or it can be either stateful or 
>> stateless? 
>> 
>> The stateless address sharing schemes (4rd and friends) allocate a port 
>> range to each subscriber statically. Static allocation is necessary for the 
>> mechanism to be stateless.
>> There are all the advantages of a stateless mechanism (see the motivation 
>> draft for a list of these advantages).
>> 
>> The problem with static allocation is that it cannot adapt quickly to 
>> subscriber demand. If for example you have one subscriber that is consuming 
>> a lot of ports, you cannot assign more ports to it. Once it's assigned, it's 
>> static and hard to change.
>> ]
>> 
>>> https://datatracker.ietf.org/doc/draft-zhou-softwire-b4-nat/
>> 
>> If I understand this correctly, it avoids NAT444 (which is a
>> good thing, and would make
>> draft-weil-shared-transition-space-request unnecessary). But
>> again, it doesn't create new ports or new public IPv4 addresses,
>> which was my point.
>> 
>> 4rd and 4via6 share this property of course.
>> 
>>   Brian
> 
> _______________________________________________
> Softwires mailing list
> [email protected]
> https://www.ietf.org/mailman/listinfo/softwires

_______________________________________________
Softwires mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/softwires

Reply via email to