RE: Consensus check

["marcelo bagnulo" <marcelo@it.uc3m.es>]

Hello Noel,

>
>     > From: "marcelo bagnulo" <marcelo@it.uc3m.es>
>
>     >> There are many different problems that could be solved under the
>     >> heading of multihoming, but for now, we should focus on site
>     >> multihoming.
>
>     > If you are asking about end-site multi-homing as oposed to provider
>     > multi-homing, i would agree.
>
> Sorry, I'm not sure I know what you mean by "end-site multihoming" and
> "provider multi-homing"; I'm assuming that the former means what
> Tony means by
> "site multi-homing", and guessing that the latter means a site that has
> multiple links to a single provider. Can you please expand on these terms?

I can try.

I think that end-sites are those that only inject packets generated by hosts
belonging to the site itself and they do not inject packets generated in
other sites in the net (the exception for this would be mutual backup
configurations).

Providers are those who inject packets that have been generated by other
sites (typically their direct or indirect clients).

I guess that these cases can be different, and that some solutions may solve
the end-site multihoming problem but are not good for solving the provider
multi-homing problem. I guess that some solutions can solve both, but i
would say that we should focus on end-site multi-homing.

>
>     > If you are asking about end-site multi-homing as oposed to host
>     > multi-homing, then:
>     > Would you include host solutions that solve the site-multihoming
>     > problems by providing a solution for each host within the site? I
>     > would say yes.
>
> There has been some reaction (e.g. from Craig H.) of the form of
> "something
> that works for a single host won't scale to large organizations with many
> thousands of hosts".

This means that this type of mechanisms are excluded, and that we should
focus in other type of mechanims?

>
> Now, it may be possible to use the same fundamental mechanism (multiple
> addresses), but layered underneath something else (e.g. 16+16, with the
> external address added at the border) which gets rid of that problem.

Agree, but other mechanisms can solve only the multi-homed host problem and
then solve the site multi-homing problem applying the host solution for
every host within the site.

>
> 	Noel
>