> >for the most part, they're only a problem if you try to use > >them in applications (where zero-configuration appliances > >are an important subset of applications) > >part of the problem is that the scope of link-local addresses > >is *not* well-defined from an application's point of view, > >since applications in general don't know, and shouldn't have > >to know, about network topology. > > as long as the applications are properly implemented with sockaddrs, > they are okay.
sockaddrs do not solve the problem of applications having to figure out how to reach a peer application in the absence of any knowledge about network topology. the source-selection rule isn't sufficient either. > the problem reside in protocols that pass IPv6 > addresses in payloads (since view of the scope is different by nodes), > including: it's entirely appropriate for applications to pass addresses in payloads, and it's essential functionality for many kinds of applications (distributed apps in particluar). for better or worse, the internet does not have a location-independent namespace that is reliable and useful for doing referrals to peer processes. and nobody has demonstrated an effective way to retrofit one into the current architecture. > we need to keep link-local IPv6 address at least for ND. use of > link-locals within zeroconf environment needs further study. we agree about that much. but they think they're ready for standards-track. Keith -------------------------------------------------------------------- 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] --------------------------------------------------------------------
