Re: (ngtrans) Joint DNSEXT & NGTRANS summary

[Keith Moore <moore@cs.utk.edu>]

>   | no you can't.  just because DNS thinks that a new prefix exists for an
>   | address is no indication that the application on the other end can
>   | deal with your using the new prefix.
> 
> No, it isn't.   But if the DNS tells you that the old address is no
> longer valid, that is an indication that you're going to have to do
> something, sometime pretty soon.

DNS cannot tell you that old addresses are no longer valid, it can only
tell you that some addresses should be valid.  The absence of some address 
in a list of A (or AAAA or A6) records is not an indication that the 
address is not valid.

>   | to do renumbering.  we need an interrupt mechanism, not a polling one,
>   | and we need something that works at the IP level, not using an application
>   | that is layered on top of IP.
> 
> Unfortunately, not possible.   IP is stateless, it has no idea who its
> peers are, to send some kind of "I just renumbered" indication (or I am
> about to renumber - same problem).  TCP could handle that, but not 
> everything uses TCP (UDP has the same problem as IP).

any system for renumbering (including the one you are proposing)
will require changes to existing protocols.  the question is, 
where is the best place to make these changes?    my guess is, 
some combination of ND/RD and ICMP.  

but they're not going to be simple changes (well, not if we expect
them to work well) because right now we don't even have the concept
of a host or stack identifier in the architecture. (with the possible
exception of the last 64 bits of a v6 address, and even then, not
in all cases) 

> Apps that need to deal with this (and that's really only those few that
> have long lived connections that need to be able to continue through events
> like renumbering ... that is, forget about HTTP, SMTP, probably FTP as well)

which apps are affected depends on the frequency of renumbering and the
overlap during which multiple addresses are valid.    and yes, if we can
make these periods large enough (and the number of apps affected few enough)
we can shift the burden of recovery to those apps.

we really need for these to be explicitly chosen design parameters.

my feeling is that the mean time between renumbering should be no worse 
than the mean time between shutdown of reasonable hosts for other reasons,
which is to say, several months.  that way, renumbering doesn't 
significantly affect application reliability.

Keith