Stephane,
Stephane Bortzmeyer wrote: Clearly, it is not so simple. While you cannot encode *everything* even in 128 bits (billing status), it is better when you can encode at least a few things.
Yes, it is better when you *can*.
Otherwise, "serious" operators would not name their routers with sensible names like P12-0.NYKCR2.New-york.opentransit.net
This takes 296 bits, btw. It is done because it *can* be done because it's a name, and that's exactly what names have been designed for. Not IP addresses. This is not the issue anyway: We are not talking about configuring the MAC address with a structure that is meaningful; this can be done already and there are 62 bits available for that. What Gert wants is a large number of bits to make the subnet number look good, in the *subnet* bits, totally ignoring conservation. Where does this come from? Nobody does that except him, nobody does it in v4, it has never been part of v6. We can't allow end-sites to do that kind of thing, they would all need at least 48 subnet bits as well. Michel.