Re: [v6ops] Implementation Status of PREF64

[Owen DeLong <[e-mail]>]

> On Sep 28, 2021, at 12:36 PM, Philip Homburg <[e-mail]> wrote:
> 
>> With SLAAC, you can begin advertising an additional prefix right
>> now.
>> 
>> You cant get rid of the old one so fast, you have to wait for the
>> timers to expire.
> 
> Indeed I was sloppy. In addition to adding a prefix right now, you can 
> also deprecate the old prefix without waiting for timers to expire.

Sort of… You can end the preferred lifetime right away and shorten the valid
lifetime significantly.

> The net effect is that new flows use the new prefix right away, but the old
> prefix can be used for 2 hours longer. So timers are limited to 2 hours.

Yes, you can shorten the valid timer to 2 hours, but it’s 2 hours from the last
time you advertised the prefix in an RA.

> If I remember correctly, there was a draft in 6man that tried to change that
> for dealing with flash renumbering. But other parts of the draft got too
> complex and the draft seems to have stalled.

It’s also a wonderful attack surface if you can forge an RA that will shut
down a network.

>> Yes, but actually, this is more effective, also as the DHCP
>> reconfigure will also be able to deprecate the previous address
>> immediately if needed.
> 
> True. Though I don't want to bet on every IoT device implementing reconfigure.

Fair enough, I wouldn’t want to, either. OTOH, I am much more willing to accept
IoT devices sticking around on the old prefix for a bit longer than most other
things.

> In any case, for a very large part of internet traffic it is important that
> new flows use the new address. Many applications agressively kill flows that
> don't seem to make progress so there no real need to wait for the old
> addresses to disappear.

Yes, that’s a common workaround for the poor real world relationship to some
aspects of the specification.

Nonetheless, there are environments where SLAAC is great and environments where
DHCPv6 is a better choice at least in the opinion of those that make the decisions
for those environments.

Lack of Android DHCPv6 support is stalling IPv6 deployment in some of those cases.

Owen