[v6ops] Re: Dynamic addresses

[Ted Lemon <[e-mail]>]

We could say that a prefix offered to a 7084 router that doesn’t have a
valid lifetime of at least double its preferred lifetime is a configuration
error, and MUST be silently discarded.

It’s still possible that someone’s game session will be interrupted, so we
could also say that the minimum valid preferred lifetime for a GUA prefix
is 24 hours, and anything shorter than that MUST be silently discarded.

It’s not clear any router vendors would adopt this behavior, but I think
it’s reasonable: might as well keep using IPv4 if they aren’t serious about
making IPv6 work.

We could also do this in hosts, but I don’t know if anyone will have the
appetite for it.


Op wo 14 aug 2024 om 10:23 schreef David Farmer <farmer=
[e-mail]>

>
>
> On Wed, Aug 14, 2024 at 04:53 Daryll Swer <contact=
> [e-mail]> wrote:
>
>> Gert Doering
>>
>> There some good arguments for dynamic prefixes
>>>
>> No there isn't.
>>
>
> All addresses are dynamic; they will change. It is a question of over what
> period and of whom or what triggers that change. Or, looking at it from
> another angle, addresses are never permanent; they will eventually change.
>
> people should stop putting explicit IPv6 addresses into places (which will
>>> inevitably cause headaches when changing ISP)
>>
>> NPTv6 to the rescue for non-renumbering across static ISP PDs that
>> changes when ISP *itself* is changed (rather than prefix within the same
>> ISP). I'm pretty confident people do not change ISPs per 6 hours or per 24
>> hours etc, as is the case with the lease time on the dynamic prefixes to
>> begin with.
>>
>
> Yes, reasonable periods of stability are necessary for both addresses and
> prefixes. All changes should be coordinated with the users or scheduled to
> occur when they cause as little disruption as possible.
>
> So actually having better support for automatic DNS registration/discovery
>>> (mDNS and friends) and just making SoHo networks renumber automatically and
>>> properly is a worthwile goal.
>>
>>
>> Not sure what you mean, mDNS and friends, does not care about changing
>> GUA or ULA prefixes, it operates on link-local addressing and does not work
>> inter-VLAN anyway (without hacking using mDNS Repeaters/Proxies because
>> link-local) — maybe I missed something here.
>>
>
> mDNS provides more than just the link-local addresses. Various DNS-SD
> proxies enable mDNS to operate beyond a single link.
>
> I do agree that those 0.001% of people that SSH into their home
>>> servers will hate SSHs dying on a renumbering event
>>
>>
>> Are you a gamer, Gert? Because many end-users are (myself included), and
>> if the games they play use IPv6 and works P2P native (we intend to have
>> IPv6 for every layer 7 application, I hope), RIP (literally) to the player,
>> when the genius ISPs flip the prefix on a live-session — same impact as the
>> SSH example.
>>
>> Do we really need to create a list of everyday applications (beyond the
>> so called 0.001% SSH users) that'd be impacted by
>> dynamic/ever-changing 24/7 prefixes? I'm sure you got the point.
>>
>
> Again, addresses and prefixes are not permanent; they inevitably change
> for one reason or another. They should not be changed haphazardly or too
> frequently, and these modifications should occur smoothly, aiming to
> minimize disruption, usually with some overlap between the old and new
> addresses or prefixes.
>
> However, the idea that addresses should not change is IPv4-centric
> thinking. IPv6 was architected with the idea that addresses are meant to be
> changed with minimal disruption. What is unacceptable is ISPs deliberately
> disrupting addresses or prefixes to differentiate service tiers; paying
> someone to prevent a disruption intentionally caused by them is considered
> extortion.
>
> Thanks
> _______________________________________________
> v6ops mailing list -- [e-mail]
> To unsubscribe send an email to [e-mail]
>