[prev in list] [next in list] [prev in thread] [next in thread] 

List:       ipng
Subject:    (IPng 4220) Re: ND: Change default prefix Lifetime from
From:       nordmark () jurassic (Erik Nordmark)
Date:       1997-07-31 0:55:01
[Download RAW message or body]


> I see. And poorly connected cable can get loose while dusted by an inept
> janitor and get reconnected after the host got kicked the next morning :)
> In any case I don't see how you solve the problem by reducing the default
> prefix lifetime to 60 days (not that I mind). 60 days is infinity for all
> practical purposes.

There is an example in the updated ND specification that
explains that infinity is a lot longer time than 60 days.

See the new Renumbering Considerations section.

> I guess, hosts that can't detect a loss of connectivity to a router within
> a reasonable period of time are troublesome. Perhaps hosts should be
> required to periodically probe routers if no other indication of
> connectivity is available and flush their prefix caches if a loss of
> connectivity is detected.

The host would detect that the router isn't present but why should
that make the host discard the autoconfigured addresses that
are still valid (and might be valid for months depending on the
lifetimes in the annoucements).

Keeping the address causes no harm as long as the host 
will not try to use the address for the unplugged network interface
as a source address when sending packets out other network
interfaces.

> Perhaps prefix caching in non-volatile memory should be disallowed for
> purpose of stateless autoconfiguration.  I.e., if a host gets a prefix and
> subsequently its address from a disk, such an address should be considered
> as statefully configured regardless of how this prefix was learned in a
> previous life.

Are you saying we should disallow hosts to have robust implementations?
Having your host be unusable for 10 minutes longer than necessary
just because your host and your router booted faster than than
your switch/hub after a power failure doesn't seem very robust to me.

DHCP allows caching of addresses with lifetimes. Why does stateless
address autoconfig have to behave worse than DHCP in this respect?

   Erik
--------------------------------------------------------------------
IETF IPng Working Group Mailing List
IPng Home Page:                      http://playground.sun.com/ipng
FTP archive:                      ftp://playground.sun.com/pub/ipng
Direct all administrative requests to majordomo@sunroof.eng.sun.com
--------------------------------------------------------------------

[prev in list] [next in list] [prev in thread] [next in thread]