RE: (multi6) requirements draft comments

["Christian Huitema" <huitema@windows.microsoft.com>]

>     >> Yes, but... if you have a solution which works for 10^7
multi-homed
>     >> houses, and works well, why bother to have a separate one for
>     >> companies?
> 
>     > If we can find a way to multihome 10^7 sites using a single
route-
> name,
>     > then I'm all for it.
> 
> ??? I never implied that any routing-based solution would allow all
10^7
> sites to use a single routing-name.

I think that when evaluating any proposal, we should try to define how
it scales by using standard complexity analysis. I can myself see at
least two interesting variables:

   1) The number S of multi-homed sites
   2) The number H of "homes" for a given site

There are multiple aspects of scaling, and we should review them. Among
the interesting variables are:

  - The impact of multi-homing on multi-homed hosts, 
  - The impact of multi-homing on "correspondent" hosts,
  - The impact of multi-homing on Internet routers, 
  - The impact of multi-homing on "site exit" routers.

Different solutions have different characteristics. The so-called "PI"
solution, for example, has an O(1) impact on multi-homed hosts and
correspondent hosts, an O(H) impact on site exit routers, and an O(S)
impact on Internet routers. The "host centric" solution that I put
forward has an O(H) impact on hosts, correspondent hosts, and site exit
routers; and the same impact on routers as standard routing, i.e. in
theory O(log S), or maybe O(log S + log H).

Expressing the complexity in these terms helps us understand the impact
of various scenarios, e.g. "what if every home network is multi-homed",
or "what if every site decides to have 100 providers"...

-- Christian Huitema