Mark Smith wrote: > > Why are you assuming an "ad-hoc network with at the extreme a couple of > hundred nodes" ?
and then wrote > I do like the idea of autoconfiguration, but in larger networks, it can > start to work against you - your network can start doing things behind > your back that make it terrible to diagnose faults. I think you just answered your own question. :) That said, I believe there are situations where local addressing is appropriate in a configured network, mainly to exploit the PI aspect of it. The difference between such networks and the ad-hoc ones I have in mind is that an architected network can partition their PI space in an aggregated manner, exactly as they would do for PA space. So requirements 2 and 3 are still applicable, but requirement 1 is not (as we both agree that someone is exercising at least some control over how the routers are configured). This still leaves the knotty problem of which address to use by default. I'm still of the opinion that the most sensible option is for the 'system' to suggest that the application by default use the smallest applicable common scope (known to the system), but make it easy for the application to override that recommendation if it thinks it knows better. There are three situations that readily come to mind where an application might want to override the default: * application uses address-based referrals. * particular addresses are known to have short lifetimes and the application plans a long-lived connection. * using link-local addresses in an environment where a device could move between links yet meaningfully keep a constant higher-level address. -- Andrew White -------------------------------------------------------------------- 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] --------------------------------------------------------------------
