Re: Ephemeral addressing [was Re: 64share v2]

[[e-mail]]

Philip,


> On 12 Nov 2020, at 11:32, Philip Homburg <[e-mail]> wrote:
> 
>> If you all you have are clients then the IPv4 Internet we have
>> today works perfectly fine.
> 
> That's not correct for quite a few reasons:
> - RIR are by and large out of IPv4 addresses. It is hard to connect a client
>  to an IPv4 internet when you have no addresses.
> - If an ISP has some IPv4 addresses then bandwidth on a CGN box is more
>  expensive than an IPv6 router
> - Servers need addresses, so we now require parties to buy second hand
>  addresses on the open market. 
> - It is perfectly possible to offer service on changing addresses. There is
>  this thing called DNS that can be used to assign a stable name to changing
>  addresses.

You seem to be making an argument for why everyone else should deploy IPv6.
Which isn't my point. My point was that IPv6 with ephemeral addressing gives equal or less benefit to the end-user than existing IPv4.
And no, there is little indication that we cannot make the IPv4 Internet work forever. If we must.

> In any case, good support for ephemeral address doesn't hurt operators who
> deply static addresses. My servers at home get static addresses from
> DHCPv4. Phones, etc. get dynamic addresses. 
> 
> Your own proposal of turning L2 bridges into L3 routers creates a lot more 
> ephemeral addresses. Currently, relative large subnets have a single /64,
> so each device that connects there will have a stable SLAAC address.
> 
> When each link has its own /64 then on wifi each time a device connects it
> will get a different address (unless 802.1X is used with radius to assign a
> static prefix to an account).

Then you didn't understand that proposal.

> So it not clear to me why you are complaining about shortlived addresses.

I'm all in favour of making renumbering work. Even to the extent of ephemeral addressing.
But glossing over the problem and pretending that it is a solved problem, is quite unfair.
Have a look at the existing work on this. E.g. RFC7010.

Best regards,
Ole