Section 2.2: Issues with IPv4 LL and Roaming Behavior
Bernard Aboba <[email protected]>
| Newsgroups | gmane.ietf.zeroconf |
|---|---|
| Message-ID | <[email protected]> |
I've noticed that portions of Section 2.2 are not very clear about the intended behavior in the case of a host changing its point of attachment. I'm concerned that the current language could do some damage -- resulting in hosts that retain LLv4 addresses across subnet moves, or even do not obtain a routable address when one was available, remaining perpetually disconnected from the Internet. 2.2. Claiming a Link-Local Address "When a network interface transitions from an inactive to an active state, the host does not have knowledge of what Link-Local IPv4 addresses may currently be in use on that link, since the network interface may have been inactive when a conflicting address was claimed. [BA] This sentence assumes that a host knows what network it is connected to when the interfaces transitions from inactive to active. When an interface transitions from inactive to active, the host not only doesn't have knowledge of LLv4 addresses in use -- it typically doesn't even know for sure whether it has connected to an alternative point of attachment. It is therefore appropriate to determine what network the host is connected to (such as via probing for a default gateway or attempting to obtain a routable address) rather than merely probing for an IPv4LL conflict, potentially on a different network than the one on which it was originally allocated. I suggest that this be changed to: "When a network interface transitions from an inactive to an active state, the host does not have knowledge of what Link-Local IPv4 addresses may currently be in use on that link, since the point of attachment may have changed or the network interface may have been inactive when a conflicting address was claimed." "Were it immediately to begin using a Link-Local IPv4 address which is already in use by another host, this would be disruptive to that other host." Worse than that -- if a host were to assume autoconf without verification whenever an interface came up, it would not obtain a routable address, even when one was available. The implication of the above two paragraphs seems to be that doing a probe first would make it ok to use an LLv4 address when an interface comes up. But this is still assuming that both LLv4 and routable address are in use on the same interface. Otherwise, this is not correct. Suggest this be changed to: "Were the host to immediately begin using a Link-Local IPv4 address which is already in use by another host, this would be disruptive to that other host. Since it is possible that the host has changed its point of attachment, a routable address may be obtainable on the new network, and therefore it cannot be assumed that a Link-Local IPv4 address is to be preferred."