Re: [v6ops] WGLC: draft-ietf-v6ops-unique-ipv6-prefix-per-host-02 - multiple prefixes per device

[Alexandre Petrescu <[e-mail]>]

Le 17/03/2017 à 14:06, Gert Doering a écrit :
> Hi,
>
> On Fri, Mar 17, 2017 at 01:05:35PM +0100, Alexandre Petrescu wrote:
>>> But seriously: how many visitors of said coffee shop would need
>>> more than a /64?
>> Many visitors use smartphone-smartwatch combinations.
>
> Those do not need multiple *networks*.

A smatphone-smartwatch combination needs one more subnet than the /64
dedicated to the smartphone by the cellular link.

>>> Yes, I've heard about the laptop with the VMs running,
>> Smartphones run VMs too, not only laptops.
>
> Really?  So these need multiple *networks* to number the VMs?

It depends on the number of VMs.  They are very easy to create, so
potentially numerous.

>>> in multiple hierarchical virtual networks, and such.  How many
>>> of those laptops exist?  100, 1000?  This is totally niche.
>>
>> Even for a 100-sized niche, one needs to plan for 100 times
>> multiple /64 prefixes per unit.
>>
>> Even for a 100-sized niche, one can't plan to use a single /64 per
>> unit.
>
> "100 world wide", and we should not engineer for those.  They will
> find ways to make their stuff work, grumbling about stupid protocols
> and everything.

The difference is whether they want something to survive the next
development cycle, be used elsewhere, etc.

>>> Normal users want their browsing, e-mail and corporate VPN to
>>> work.
>>
>> New ways dont do any browsing, email or corporate VPN.  They do
>> purpose-specific messaging and project management apps.  Often they
>>  neeed simultaneity between multiple independent such apps - easier
>> with VMs.
>
> All that is "talk to internet from clients", and a whole /64 per
> device would be extremely plenty for that.

This "talk to internet from clients" exclusive perspective makes remote
software update and other low-latency "ring" notifications impossible,
or simply too CPU-consuming.

> You need more than a /64 if you have routed topologies, which you
> just do not *have* inside your smartwatch.

I didnt mean inside the smartwatch.

I meant the smartphone does route, and the smartphone-smartwatch forms a
subnet that needs a /64 for itself (in addition to the /64 presumably
allocated to the smartphone).

Alex

>
> Gert Doering -- NetMaster
>