Bob, > >=> My 2 cents on the preference values: > >It's not only that there is only 3 values, but > >that these values may be used across different > >administrative domains (multihomed hosts). They > >need to be consistently defined so 'good' in > >ISP X domain is roughly the same as 'good' > >in ISP Y's domain. How can we do that ? > >Maybe some concrete definitions ? > > It's a hard problem. Routing has exactly the same problem > and I don't > think they ever came up with an automatic way to deal with > this. The best > that may be possible is to have a manual mapping between > domain X and > domain Y. This was one of the reasons why there is a > separation between > intra-domain and inter-domain routing. There wasn't a good > way to do an > automatic mapping between metrics from one domain to another.
=> I agree with your analysis and conclusion with one clarification: Two domains may decide to agree on certain mappings as long as they have control over the devices that receive this information. I.e. router_x and router_y. However, in this case we have a scenario where a 'host' is receiving the equivalent of link state information from 2 different routers that happened to be geographically adjacent. The 2 different domains may not even be aware of each others existence, let alone have some agreement. This is obviously a wireless case. I haven't been giving a lot of thought to a solution, but I can think of 2 possible options: 1.Have concrete definition of preferences and increase the ranges (from 3 to 100?) so that a host can roughly compare between two preferences from 2 routers in different domains. 2. Specify clearly in the draft that this solution is only useful for a host connected to 2 routers in the same administrative domain. At least in this case the host would know whether it should rely on the preference values or ignore them (presumably the host is aware that it's connected to different admin domains...somehow) 3. Have some way of showing the host that the two domains are in agreement about the meaning of the preferences. I'm leaning towards 2 above. I think it's the easiest way. Anyhow, other groups (seamoby) are considering capability discovery for routers and I suspect that there will be many more parameters to be added. To me, 'preference' seems to aggregate several parameters (huge!) without explicityly stating them and the weight of each parameter. Cheers, Hesham -------------------------------------------------------------------- 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] --------------------------------------------------------------------
