Re: LL70: Use of the term "valid"

Philip Nye said:

>I think that the use of "valid" conveys meaning because it helps people
>grasp the idea that a host may have been configured with an address which
>was previously valid but although still configured, is no longer useable
>because the network has changed (e.g. the host moved to a new network).
>This is a common case and there is often good reason not to simply
>unconfigure that address because the host may soon be moved back to a
>context in which it is valid - meanwhile assigning an IPv4LL address
>provides some connectivity.

I think there is a distinction between "configured" and "useable".  I'm
not entirely clear what "useable" means here -- is this an address that is
configured, but might be only useable for receiving but not sending
packets?  If so, then it might make more sense to talk about "active" or
"deprecated" addresses than "valid" ones.

>The validity of a configured address is difficult to establish because
>the context in which it is intended to operate or becomes invalid can
>vary.  Therefore I think that all we can do is explain the concept and
>present the DNAv4 reachability test as a useful example (informative) rather than
>a definition (normative) of validity.

[DNAv4] uses the reachability test to determine whether to do DHCPv4 or
not, not to decide "validity" as discussed here.  It could be that a
configured IPv4 routable address fails the reachability test, because
the default gateway failed, but then a subsequent DHCPDISCOVER obtains
the same IP address, but a different default gateway.  In this case
configuring an IPv4 LL address in between the reachability test and the
DHCP conversation would be a bad idea.

I'm somewhat uncomfortable with where the spec is going with some of these
considerations because while it is citing [DNAv4], the implied behavior is
not in sync with that document.