Re: [Autoconf] Autoconf addressing model

[Alexandru Petrescu <[e-mail]>]

Teco, the problem is simple: host-based routes are not preferrable.

Host-based routes in routers surrounding router X (this is what a /128
address on a virtual interface on router X implies, be that loopback
interface or not loopback interface) for a dynamically changing topology
may lead to too many routes coming up and down, too many specific
entries in the routing tables, too much dedicated routing messaging, and
more.

That only for the network itself.

Connecting it to the Internet, with host-based routes propagated up and
down throughout, may risk influencing the routes in the Internet proper.
  A new host-based route inserted in the routing table of the gateway
connecting this network to the Internet may propagate throughout the
entire fixed access system.

If you told me which part of the above were unclear I could explain
further.  But I think there's large agreement on it.

About the feasibility of _one_ /128 address on 'lo' while using _one_
host-based route: as you pointed out with the ios-linux example, I fully
agree a /128 address assigned on the 'lo' interface of one computer, and
a host-based route in the other computer towards that /128 address -
will work.  But just one.

About posting to v6ops: which part do you think I should post there?

About singling me out being the first and only mentioning some problem:
I can only be happy about it and stop posting :-)  But this host-based
route stuff is not new, I'm myself surprised I'm singled out about it.

Alex

Teco Boot a écrit :
> Hi Alex,
> 
> As far as I know, you are the first person that came up with problems
>  using a loopback interface for globally unique addresses / host 
> prefixes (/128, /32) for routers. Please provide good argumentation,
>  otherwise we follow the already accepted practice in the routing 
> community, also documented in RFC5375 (and others, e.g. RFC3484).
> 
> Maybe you should post this in v6ops, not in Autoconf.
> 
> Teco.
> 
> 
> 
>