[arin-ppml] Draft Policy ARIN-2019-2: Waiting List Block Size Restriction

[Owen DeLong]

> On Mar 2, 2019, at 13:11 , Jo Rhett <jrhett at netconsonance.com> wrote:
> 
>> I mean what do people even need lots of IPv4 for anymore anyway?  A single
>> web server with a single IPv4 address can easily support tens of thousands
>> of distinct and unique web sites.  A single mail server on a single IPv4
>> address can likewise support mail services for tens of thousands of
>> recipient and sender domain names.  A single name server on a single IPv4
>> address can also provide DNS service for tens of thusands of domain names.
> 
> You aren’t taking into account limitations of IPv4 64k ports and reuse timers. This protocol was created in the 1970s, and port reuse has been sped up but cannot be solved.

Yes and no… Remember, that’s 64k ports per remote IP + Port combination… It’s a four element tuple (local IP, local port, remote IP, remote port) and only one element needs to be unique to make a unique session. So theoretically, even between two hosts, you have 4 billion possible sessions. Obviously, that’s reduced if you’re limited to a well known listener port at one end (the usual case), but you still have literally 64512 sessions per remote IP possible to each well known port, assuming we exclude the”privileged” ports on the client system.

> 
>> For anyone needing to support big batches of end-luser clients, there is IPv6.
> 
> 
> While I wholeheartedly agree with that sentiment, there are shocking amounts of US-based enterprises where IPv6 is still not available. This means that anyone offering services to the enterprise requires IPv4.

Or it means it’s time for those enterprises to get with the program. IPv4 isn’t sustainable no matter what we do or don’t do about these particular relatively small batches of addresses.

Owen