On Aug 28, 2003, at 4:25 PM, Erik Nordmark wrote:
The conflict resolution protocol is just a way of letting the user know
that a particular name is ambiguous in the domain that is currently
reachable. One of the users will then have to pick another name and
announce it to any other users who knew about the old name and were
using it. Having the conlict resolution protocol remove a name or change
it to something else will not remove the confusion, since the old name
will still be lodged into the minds of the people who need it.
Frankly, I don't see that having multiple intefaces makes all that much
difference.
The difference is that the names need to be unique not only on one link,
but across all interfaces connected to the multihomed host.
Do they really? If so, how do we guarantee this? As you mentioned, some sort of conflict resolution protocol would have to allow a node straddling two networks notify nodes of conflicts. This is a very difficult problem to solve. For example, if there are two machines that are connect to the same wired network and wireless network, machine A may see responses from machine B on the wireless network as conflicting with the responses on the wired network from machine B. Machine A has no way to know that the responses are both from the exact same machine so this is perfectly valid.
How would you expect conflict resolution to handle that?
I don't think this is something we can solve. I also don't think it is something particularly important to solve. There are a number of simple solutions that could work well. For example, if it was possible to append a name with the interface the query should be sent on (i.e. laptop%en0), this would give someone the ability to specify which "laptop" they were referring to in the case that just specifying laptop would not be enough. Specifying the name alone will often work in 99% of cases. Adding additional functionality to allow this to work for the additional 1% of the cases might be worthwhile. If requiring the protocol perfectly handle that last 1% (the edge cases) results in an impossible task, is it really worth throwing out the protocol that works for 99% of the cases just because we can't perfectly solve the edge case?
-josh
-------------------------------------------------------------------- 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 [EMAIL PROTECTED] --------------------------------------------------------------------
